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
.
2
,
A
p
r
il
20
25
,
p
p
.
1
5
7
2
~
1
5
8
3
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
2
.
pp
1
5
7
2
-
1
5
8
3
1572
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
No
v
el
features
ext
ra
ction: pigm
en
t
epithelial d
etach
ment
detec
tion usin
g
m
a
chine learning
al
g
o
rithms
Sh
ee
ba
T
ha
nk
a
pp
a
n
M
er
cy
1
,
Albert
Ant
o
ny
Ra
j
Sa
m
ina
t
ha
n
1
,
Ana
nd
M
urug
a
dh
a
s
2
,
Ans
hy
P
rincel
la
Ana
nd
Sh
ee
ba
3
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
s,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
H
u
ma
n
i
t
i
e
s,
S
R
M
I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
,
K
a
t
t
a
n
k
u
l
a
t
h
u
r
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
N
e
t
w
o
r
k
i
n
g
a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
s
,
S
R
M
I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
,
K
a
t
t
a
n
k
u
l
a
t
h
u
r
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
S
R
M
I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
,
K
a
t
t
a
n
k
u
l
a
t
h
u
r
,
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
Ma
r
1
8
,
2
0
2
4
R
ev
is
ed
Oct
1
,
2
0
2
4
Acc
ep
ted
Oct
2
3
,
2
0
2
4
Th
e
m
a
jo
rit
y
o
f
t
h
e
re
ti
n
a
l
d
ise
a
se
s
h
a
v
e
v
isu
a
l
sy
m
p
to
m
s.
An
y
a
re
a
o
f
th
e
re
ti
n
a
,
a
d
e
li
c
a
te
lay
e
r
o
f
t
issu
e
o
n
t
h
e
i
n
terio
r
th
e
p
o
ste
rio
r
w
a
ll
o
f
th
e
h
u
m
a
n
e
y
e
,
c
a
n
b
e
imp
a
c
ted
b
y
re
ti
n
a
l
d
iso
rd
e
rs.
Op
ti
c
a
l
c
o
h
e
re
n
c
e
to
m
o
g
ra
p
h
y
(OCT)
is
th
e
u
tmo
st
c
o
m
m
o
n
ly
u
se
d
ima
g
i
n
g
p
ro
c
e
d
u
re
fo
r
d
iag
n
o
sin
g
re
ti
n
a
l
d
is
o
rd
e
rs
su
c
h
a
s
a
g
e
-
re
late
d
m
a
c
u
lar
d
e
g
e
n
e
ra
ti
o
n
(ARMD),
d
iab
e
ti
c
re
ti
n
o
p
a
th
y
,
p
i
g
m
e
n
t
e
p
it
h
e
li
a
l
d
e
tac
h
m
e
n
t
(P
ED),
m
a
c
u
lar
h
o
les
,
a
n
d
m
o
re
.
In
t
h
is
stu
d
y
,
we
p
u
t
fo
r
th
a
b
ra
n
d
-
n
e
w
tec
h
n
iq
u
e
fo
r
a
c
c
u
ra
tely
e
x
trac
ti
n
g
fe
a
tu
r
e
s
fro
m
OCT
ima
g
e
s
to
id
e
n
ti
fy
P
E
D
d
ise
a
se
s.
F
o
r
t
h
e
p
re
p
ro
c
e
ss
in
g
ste
p
,
we
e
x
a
m
in
e
d
t
h
e
wie
n
e
r
fil
terin
g
m
e
th
o
d
.
Af
ter t
h
a
t,
we
se
g
m
e
n
ted
th
e
re
ti
n
a
l
p
i
g
m
e
n
t
e
p
it
h
e
li
u
m
(
R
P
E
)
lay
e
r
u
se
d
to
th
e
t
h
re
sh
o
l
d
in
g
m
e
th
o
d
,
e
x
trac
ted
th
e
fe
a
tu
re
s
fro
m
t
h
e
RP
E
lay
e
r
,
a
n
d
th
e
n
g
a
v
e
t
h
e
fe
a
tu
re
s
to
m
a
c
h
in
e
lea
rn
in
g
(
M
L)
c
las
sifiers
li
k
e
th
e
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
(S
VM),
l
o
g
isti
c
re
g
re
ss
io
n
(LR)
,
k
-
n
e
a
re
st
n
e
i
g
h
b
o
r
s
(KN
N),
d
e
c
isio
n
tree
(DT),
ra
n
d
o
m
fo
re
st
(RF
),
n
a
iv
e
Ba
y
e
s
(
NB),
a
n
d
a
rti
ficia
l
n
e
u
ra
l
n
e
two
r
k
(AN
N).
Th
e
t
o
tal
d
a
tas
e
t
a
b
o
u
t
2
0
0
ima
g
e
s
a
m
o
n
g
1
0
0
is
n
o
rm
a
l
a
n
d
1
0
0
is
P
ED,
we
train
e
d
t
h
e
d
a
tas
e
t
a
s
a
n
u
n
b
a
lan
c
e
d
a
n
d
b
a
lan
c
e
d
g
r
o
u
p
.
Th
e
RF
is
t
h
e
b
e
st
o
u
tco
m
e
in
c
o
m
p
a
riso
n
o
f
o
th
e
r
c
las
sifiers
.
Th
e
o
v
e
ra
ll
o
u
tc
o
m
e
o
f
ra
n
d
o
m
f
o
re
st
is
1
0
0
%
a
c
c
u
ra
c
y
.
K
ey
w
o
r
d
s
:
Ar
tific
ial
n
eu
r
al
n
etwo
r
k
Ma
ch
in
e
lear
n
in
g
Op
tical
co
h
er
en
ce
to
m
o
g
r
ap
h
y
Pig
m
en
t e
p
ith
elial
d
etac
h
m
en
t
Pig
m
en
t e
p
ith
eliu
m
lay
e
r
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
:
An
an
d
Mu
r
u
g
ad
h
as
Dep
ar
tm
en
t o
f
Netwo
r
k
in
g
an
d
C
o
m
m
u
n
icatio
n
s
,
SR
M
I
n
s
titu
te
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
Kattan
k
u
lath
u
r
-
6
0
3
2
0
3
,
T
am
il Na
d
u
,
I
n
d
ia
E
m
ail: a
n
an
d
m
3
@
s
r
m
is
t.e
d
u
.
i
n
1.
I
NT
RO
D
UCT
I
O
N
A
s
p
ec
ialized
s
en
s
e
o
r
g
an
c
alled
th
e
h
u
m
a
n
e
y
e
is
a
b
le
to
ta
k
e
in
v
is
u
al
im
ag
es,
wh
ich
ar
e
s
u
b
s
eq
u
en
tly
tr
an
s
m
itted
to
th
e
b
r
ain
.
I
t
h
as
a
s
p
h
er
e
-
lik
e
s
h
ap
e
an
d
r
esp
o
n
d
s
to
lig
h
t
to
g
iv
e
h
u
m
an
s
th
e
ab
ilit
y
to
s
ee
.
T
h
e
b
eh
in
d
s
e
g
m
en
t
o
f
th
e
v
is
u
al
o
r
g
a
n
is
p
u
t
u
p
o
f
th
e
v
itre
o
u
s
b
o
d
y
,
r
etin
a,
ch
o
r
o
id
,
an
d
s
cler
a,
wh
ile
th
e
f
o
r
em
o
s
t
s
eg
m
en
t
o
f
th
e
ey
e
is
p
u
t
u
p
o
f
t
h
e
co
r
n
ea
,
ir
is
,
an
d
le
n
s
.
J
u
s
t
b
etwe
en
th
e
ir
is
an
d
p
u
p
il,
th
er
e
is
a
len
s
th
at
s
u
p
p
o
r
ts
b
r
ig
h
ten
th
e
p
o
s
ter
io
r
p
l
an
e
o
f
th
e
ey
e.
T
h
e
m
ajo
r
o
f
th
e
v
is
u
al
o
r
g
an
is
o
cc
u
p
ied
with
th
e
v
itre
o
u
s
,
a
tr
an
s
lu
ce
n
t
g
el
[
1
]
.
E
y
e
p
u
p
il
an
d
len
s
let
th
in
-
lig
h
t
to
r
ea
ch
th
e
in
n
er
ey
e
s
u
r
f
ac
e.
T
h
e
in
n
er
ey
e
s
u
r
f
ac
e
is
a
lay
er
o
f
u
n
i
q
u
e
lig
h
t
-
s
en
s
in
g
ce
lls
th
at
co
v
er
s
th
e
i
n
s
id
e
o
f
th
e
ey
e.
I
t
co
n
v
er
ts
lig
h
t
in
t
o
elec
tr
ical
im
p
u
ls
e.
T
h
ese
im
p
u
ls
es
tr
av
e
l
f
r
o
m
th
e
r
etin
a
to
t
h
e
b
r
ain
b
y
th
e
o
p
tic
-
n
er
v
e.
T
h
e
m
ac
u
la,
a
tin
y
,
p
ar
ticu
lar
l
y
s
en
s
itiv
e
ar
ea
o
f
th
e
r
etin
a,
i
s
r
esp
o
n
s
ib
le
f
o
r
th
e
ce
n
te
r
v
i
s
io
n
.
T
h
e
ty
p
e
a
n
d
am
o
u
n
t
o
f
p
ig
m
en
t
f
o
u
n
d
in
t
h
e
ir
is
d
ef
in
es
th
e
co
lo
r
o
f
th
e
ey
es.
A
p
er
s
o
n
'
s
ey
e
co
lo
r
is
d
eter
m
in
ed
b
y
h
o
w
m
an
y
g
en
es
th
ey
r
ec
ei
v
e
f
r
o
m
ea
ch
p
ar
en
t
[
2
]
.
Fig
u
r
e
1
d
ep
icts
th
e
h
u
m
an
ey
e
an
ato
m
y
.
E
v
er
y
d
ay
,
r
etin
a
l
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
N
o
ve
l fe
a
tu
r
es e
xtra
ctio
n
:
p
ig
men
t e
p
ith
elia
l d
eta
ch
me
n
t
d
e
tectio
n
u
s
in
g
…
(
S
h
ee
b
a
Th
a
n
ka
p
p
a
n
Mer
cy
)
1573
an
o
m
alies
ar
e
d
iag
n
o
s
ed
in
th
o
u
s
an
d
s
o
f
p
e
o
p
le.
Usi
n
g
o
p
tical
co
h
er
en
ce
to
m
o
g
r
ap
h
y
(
OC
T
)
s
ca
n
im
ag
es,
q
u
alif
ied
m
e
d
ical
p
er
s
o
n
n
el
m
u
s
t
in
s
p
ec
t
an
d
ca
teg
o
r
ize
r
etin
al
ab
n
o
r
m
alities
as
p
ar
t
o
f
t
h
e
d
ia
g
n
o
s
in
g
p
r
o
ce
d
u
r
e.
I
n
g
en
e
r
al,
th
is
is
a
d
r
awn
-
o
u
t
m
an
u
al
p
r
o
ce
s
s
.
T
h
e
d
etec
tio
n
an
d
class
if
icatio
n
o
f
m
ed
ical
i
m
ag
es
u
s
in
g
m
ac
h
in
e
in
tellig
en
ce
is
n
o
w
m
o
r
e
p
r
ac
tical
b
ec
au
s
e
to
r
ec
en
t
d
ev
elo
p
m
en
ts
i
n
co
m
p
u
ter
v
is
io
n
.
Alter
n
ativ
e
s
o
lu
tio
n
s
to
th
is
i
s
s
u
e,
s
u
ch
as
th
e
cr
ea
tio
n
o
f
an
au
to
m
ated
p
r
o
g
r
am
th
at
m
ig
h
t
h
elp
m
ed
ical
p
er
s
o
n
n
el
in
th
e
d
etec
tio
n
an
d
ca
teg
o
r
izatio
n
p
r
o
ce
s
s
ea
r
lier
,
co
u
ld
b
e
f
o
u
n
d
.
Fig
u
r
e
1
.
Hu
m
an
ey
e
Usi
n
g
OC
T
,
is
a
s
im
p
le
s
ca
n
n
in
g
ap
p
r
o
ac
h
th
at
g
en
er
ates
h
eig
h
ten
ed
cla
r
ity
im
ag
es
a
cr
o
s
s
-
s
ec
tio
n
in
r
etin
a
im
ag
es,
p
h
y
s
ician
s
ca
n
co
llect
v
alu
ab
le
in
f
o
r
m
atio
n
f
o
r
id
en
tify
in
g
m
an
y
r
etin
al
d
is
ea
s
es
.
Desp
ite
i
ts
ad
v
an
tag
es,
m
an
u
al
OC
T
im
a
g
e
an
aly
s
is
n
ee
d
s
ex
ten
s
iv
e
ti
m
e
an
d
is
g
r
ea
tly
r
elian
t
o
n
th
e
an
aly
s
t's
p
er
s
o
n
al
Pro
f
icien
cy
.
T
h
e
s
tu
d
y
[
3
]
f
o
c
u
s
es
o
n
th
e
clin
ical
in
ter
p
r
etatio
n
o
f
r
etin
al
d
is
o
r
d
er
s
u
s
in
g
m
ac
h
in
e
lear
n
in
g
to
ex
am
in
e
OC
T
im
ag
es.
T
h
e
s
am
p
le
OC
T
im
ag
e
is
s
h
o
wn
in
Fig
u
r
e
2
.
T
h
e
r
e
ar
e
1
1
lay
er
s
in
th
e
h
u
m
a
n
ey
e.
Fig
u
r
e
3
s
h
o
ws
th
e
r
etin
al
lay
er
s
in
th
e
h
u
m
an
r
etin
al
OC
T
im
ag
es.
T
h
e
lay
e
r
s
in
clu
d
e
:
i
n
n
er
p
h
o
to
r
ec
ep
to
r
lay
er
,
in
n
e
r
n
u
clea
r
lay
er
,
in
n
er
p
lex
if
o
r
m
lay
er
,
ch
o
r
o
id
,
o
u
ter
n
u
clea
r
la
y
er
,
o
u
ter
p
le
x
i
f
o
r
m
lay
e
r
,
r
etin
al
p
ig
m
en
t e
p
ith
eliu
m
lay
er
,
an
d
ex
ter
n
al
lim
itin
g
m
em
b
r
an
e.
Fig
u
r
e
2
.
No
r
m
al
OC
T
im
ag
es
Fig
u
r
e
3
.
R
etin
al
lay
er
s
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
.
2
,
Ap
r
il
20
25
:
1
5
7
2
-
1
5
8
3
1574
Ag
e
-
r
elate
d
m
ac
u
lar
d
e
g
en
er
a
tio
n
(
AM
D)
,
th
e
m
o
s
tly
f
r
eq
u
en
t
ey
e
d
is
ea
s
e
p
r
o
ce
s
s
,
is
ch
a
r
ac
ter
ized
b
y
a
f
r
eq
u
e
n
t
ch
ar
ac
te
r
is
tic
ca
lled
r
etin
al
p
ig
m
e
n
t
ep
ith
e
lial
d
etac
h
m
en
t
(
PED)
[
4
]
.
I
n
itial
r
esear
ch
was
r
estricte
d
to
e
v
alu
atin
g
th
ese
PEDs
u
s
in
g
p
lan
a
r
im
a
g
in
g
te
ch
n
iq
u
es
s
u
c
h
f
lu
o
r
escein
an
g
io
g
r
ap
h
y
a
n
d
co
lo
r
f
u
n
d
u
s
p
h
o
to
g
r
a
p
h
y
.
T
h
e
v
is
u
al
s
y
s
tem
o
f
f
er
s
a
u
n
i
q
u
e
win
d
o
w
in
to
th
e
a
g
ein
g
p
r
o
ce
s
s
an
d
th
e
ch
allen
g
es
ass
o
ciate
d
with
d
ev
elo
p
i
n
g
ef
f
ec
tiv
e
tr
ea
tm
en
ts
f
o
r
a
g
e
-
r
ela
ted
ey
e
illn
ess
es.
T
h
is
ar
ticle
s
u
g
g
ests
a
n
u
m
b
er
o
f
p
o
ten
tial
ML
tech
n
iq
u
es
to
tack
le
th
ese
is
s
u
es.
R
etin
al
im
ag
e
s
cr
ee
n
in
g
was
em
p
lo
y
ed
as
an
ex
am
p
le
ap
p
licatio
n
-
d
o
m
ain
to
s
er
v
e
a
s
th
e
in
v
esti
g
atio
n
'
s
f
o
ca
l
p
o
in
t
an
d
to
ass
ess
th
e
p
r
o
v
id
ed
th
eo
r
ies
[
5
]
.
Mo
r
e
s
p
ec
if
ically
,
AM
D
was
d
etec
ted
b
y
class
if
y
in
g
im
a
g
es
as
AM
D
n
o
r
m
al/ab
n
o
r
m
al
illn
ess
.
Fig
u
r
e
4
s
h
o
ws
th
e
PED
im
ag
es.
T
h
e
r
etin
a
is
th
e
in
n
er
lay
e
r
o
f
th
e
ey
e
f
r
o
m
wh
ich
lig
h
t
s
ig
n
als
ar
e
tr
an
s
f
o
r
m
e
d
in
to
n
eu
r
o
lo
g
ical
s
ig
n
als.
Sin
ce
it
co
n
tain
s
a
v
ar
iety
o
f
r
o
d
a
n
d
co
n
e
p
ig
m
en
ts
,
it
is
ce
r
tain
to
d
u
ll
lig
h
t
an
d
af
f
ec
t
co
lo
r
p
e
r
ce
p
tio
n
.
T
h
er
e
ar
e
s
ev
er
al
lay
er
s
o
f
s
an
itar
y
an
d
p
h
y
s
io
lo
g
ical
c
o
n
ce
r
n
in
th
e
r
etin
a.
Gen
er
ally
s
p
ea
k
in
g
,
d
a
m
ag
e
to
s
p
ec
if
ic
lay
er
s
ca
u
s
es
v
ar
io
u
s
ab
n
o
r
m
alities
as
wel
l
as
v
is
u
al
lo
s
s
.
Fo
r
th
e
d
iag
n
o
s
is
o
f
s
ev
er
al
r
etin
al
illn
ess
e
s
,
in
clu
d
in
g
ce
n
tr
al
-
s
er
o
u
s
-
r
etin
o
p
at
h
y
,
d
iab
etic
-
r
etin
o
p
ath
y
,
m
ac
u
lar
-
ed
em
a,
m
ac
u
la
r
-
h
o
le,
ch
r
o
n
ic
g
lau
co
m
a,
an
d
b
lin
d
n
ess
,
OC
T
r
em
ain
s
a
cr
u
cial
to
o
l.
Xian
g
et
a
l.
[
6
]
i
n
tr
o
d
u
ce
d
th
e
f
lu
id
s
eg
m
en
tatio
n
tech
n
i
q
u
e
f
o
r
OC
T
r
etin
al
im
ag
es.
Xian
g
e
t
a
l.
[
7
]
d
ep
lo
y
ed
a
s
u
p
er
v
is
ed
m
ac
h
in
e
lea
r
n
in
g
s
tr
ateg
y
to
s
u
g
g
est
s
eg
m
en
tin
g
lay
er
s
an
d
n
eo
v
ascu
lar
izatio
n
u
s
in
g
OC
T
im
ag
es.
I
n
s
tu
d
y
[
8
]
,
to
o
r
g
a
n
ize
s
u
b
r
etin
al
f
lu
i
d
(
SR
F)
an
d
P
E
D
asp
ec
ts
am
o
n
g
OC
T
im
a
g
es.
In
s
tu
d
y
[
9
]
,
o
p
tical
co
h
er
en
ce
to
m
o
g
r
a
p
h
y
(
OC
T
)
im
ag
es
ca
n
b
e
u
tili
ze
d
to
r
ec
o
g
n
ize
th
e
m
ac
u
lar
h
o
l
e
d
escr
ib
ed
b
y
Me
n
d
es
et
a
l.
th
r
o
u
g
h
th
e
u
p
p
e
r
b
o
r
d
er
s
o
f
th
e
in
ter
n
al
lim
itin
g
m
em
b
r
an
e
(
I
L
M)
a
n
d
t
h
e
r
etin
al
p
ig
m
en
t
e
p
ith
eliu
m
(
R
PE)
.
Aza
m
en
et
a
l.
[
1
0
]
s
u
g
g
ested
em
p
lo
y
in
g
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
b
ased
o
n
r
is
k
v
ar
iab
les
to
f
o
r
ec
ast
th
e
o
n
s
et
o
f
d
iab
etic
r
etin
o
p
at
h
y
.
Fig
u
r
e
4
.
Sam
p
le
PED
im
a
g
es
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
A
s
ep
ar
atin
g
o
f
th
e
R
PE
f
r
o
m
th
e
in
s
id
e
co
llag
en
ic
la
y
er
o
f
B
r
u
c
h
-
m
em
b
r
an
e
is
ass
ig
n
ed
to
as
r
etin
al
p
ig
m
en
t
e
p
ith
elial
d
etac
h
m
en
t
(
PED)
.
B
o
th
th
e
d
r
y
an
d
wet
f
o
r
m
s
o
f
ag
e
-
r
elate
d
-
m
ac
u
la
r
-
d
eg
en
er
atio
n
f
r
e
q
u
en
tly
e
x
h
i
b
it
th
is
s
y
m
p
to
m
.
T
o
b
e
ab
l
e
to
aid
in
th
e
id
e
n
tify
in
g
a
n
d
m
an
a
g
in
g
o
f
th
is
s
ig
n
if
ican
t
asp
ec
t
o
f
ag
e
-
r
elate
d
m
ac
u
lar
d
e
g
en
er
atio
n
(
AR
MD
)
,
th
is
ex
am
in
atio
n
p
u
r
p
o
s
es
to
g
iv
e
a
th
o
r
o
u
g
h
o
v
e
r
v
iew
o
f
t
h
e
p
ath
o
lo
g
y
,
m
e
d
ical
an
d
im
a
g
in
g
asp
ec
ts
,
n
o
r
m
al
f
lo
w
a
n
d
t
h
e
r
ap
y
o
f
th
e
s
ev
er
al
f
o
r
m
s
o
f
PED
[
1
1
]
–
[
1
3
]
.
K
o
h
et
a
l.
[
1
4
]
em
p
lo
y
ed
4
0
4
n
o
r
m
al
an
d
1
,
0
8
2
p
ath
o
lo
g
ical
f
u
n
d
u
s
r
etin
al
p
ictu
r
es
to
class
if
y
r
eg
u
lar
f
r
o
m
ir
r
e
g
u
lar
v
is
u
als
u
s
in
g
a
r
an
d
o
m
f
o
r
est
(
R
F)
clas
s
if
ier
.
W
i
th
o
u
t
s
eg
m
en
tin
g
,
en
er
g
y
an
d
en
tr
o
p
y
in
f
o
r
m
atio
n
wer
e
ex
tr
ac
ted
u
s
in
g
th
e
2
D
co
n
ti
n
u
o
u
s
wav
elet
tr
an
s
f
o
r
m
,
wh
ich
p
r
o
d
u
ce
d
g
r
ea
t
ac
cu
r
ac
y
an
d
allo
wed
th
e
clas
s
if
icatio
n
o
f
th
r
ee
d
is
tin
ct
g
r
o
u
p
s
o
f
f
u
n
d
u
s
im
ag
es.
I
n
o
r
d
e
r
to
id
en
tify
p
h
o
to
s
with
an
d
with
o
u
t
f
at
d
ep
o
s
its
,
s
u
p
p
o
r
t
v
ec
t
o
r
m
ac
h
in
e
(
S
VM
)
,
k
-
n
ea
r
est
n
ei
g
h
b
o
r
s
(
K
NN
)
,
an
d
m
ac
h
in
e
lear
n
in
g
(
ML
)
class
if
icatio
n
wer
e
ex
am
in
ed
,
wh
ich
p
r
o
d
u
ce
d
g
r
ea
ter
ac
cu
r
ac
y
an
d
s
en
s
itiv
ity
[
1
4
]
.
T
h
e
th
r
ee
f
ea
tu
r
e
ex
tr
ac
tio
n
tec
h
n
iq
u
es
wer
e
co
m
p
ar
e
d
,
an
d
in
a
d
d
in
g
to
d
if
f
er
en
tiatin
g
c
h
ar
ac
ter
is
tics
,
h
id
d
en
Ma
r
k
o
v
m
o
d
el
(
HM
M)
an
d
SVM
class
if
ier
s
wer
e
em
p
lo
y
ed
to
d
e
ter
m
in
e
wh
eth
er
th
e
in
d
icato
r
was
Par
k
in
s
o
n
'
s
-
r
elate
d
o
r
n
o
t
[
1
5
]
.
Me
d
ical
im
ag
e
co
m
p
r
ess
io
n
is
ac
co
m
p
lis
h
ed
u
s
in
g
r
an
d
o
m
p
er
m
u
tatio
n
co
ef
f
icie
n
t
-
b
ase
d
co
m
p
r
ess
ed
s
en
s
in
g
(
C
S).
W
ith
th
is
tech
n
iq
u
e,
a
h
ig
h
-
q
u
ality
f
latten
e
d
im
ag
e
is
m
ain
tain
ed
with
a
r
esp
ec
tab
le
am
o
u
n
t
o
f
t
h
e
r
es
to
r
ed
im
a
g
e
[
1
6
]
.
T
h
e
am
p
litu
d
e
m
o
d
u
latio
n
-
f
r
e
q
u
en
c
y
m
o
d
u
latio
n
(
AM
-
FM)
tech
n
o
lo
g
y
is
u
s
ed
to
an
aly
ze
tex
tu
r
es
in
s
eg
m
en
ted
im
ag
es.
E
ac
h
im
ag
e
s
am
p
le'
s
cu
r
r
en
t
am
p
litu
d
e
an
d
f
r
eq
u
e
n
cy
d
ata
ar
e
g
ath
er
e
d
an
d
q
u
an
tifie
d
.
T
o
ass
ess
if
th
e
ar
ter
y
is
n
o
r
m
al
o
r
ab
er
r
a
n
t,
it
is
th
en
as
s
o
ciate
d
to
a
ty
p
ical
m
ater
ial
m
o
d
el
[
1
7
]
.
ML
s
y
s
tem
wr
itten
in
MA
T
L
AB
was
u
s
ed
to
id
en
tify
AM
D
cr
ea
ted
o
n
f
u
n
d
u
s
r
etin
al
p
ictu
r
es
f
r
o
m
th
e
FR
I
d
atab
ase.
C
o
m
p
ar
ed
to
o
th
er
ML
class
if
ier
s
,
th
e
SVM
h
as
a
class
if
icatio
n
ac
cu
r
ac
y
th
at
i
s
h
ig
h
er
(
>9
3
%)
[
1
8
]
.
L
ay
e
r
ex
tr
ac
tio
n
o
f
OC
T
im
ag
es
h
as
b
ee
n
ad
v
an
ce
d
th
r
o
u
g
h
a
v
ar
iety
o
f
tech
n
iq
u
es,
with
th
e
g
r
ap
h
s
ea
r
c
h
alg
o
r
ith
m
s
er
v
i
n
g
as
th
e
co
r
n
er
s
to
n
e
in
m
an
y
ap
p
r
o
ac
h
es
[
1
9
]
–
[
2
6
]
.
R
ec
en
tl
y
,
T
ian
et
a
l.
[
2
7
]
s
u
g
g
ested
a
s
h
o
r
test
-
p
ath
d
ep
e
n
d
en
t
g
r
ap
h
-
s
ea
r
ch
to
d
e
d
u
ct
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
N
o
ve
l fe
a
tu
r
es e
xtra
ctio
n
:
p
ig
men
t e
p
ith
elia
l d
eta
ch
me
n
t
d
e
tectio
n
u
s
in
g
…
(
S
h
ee
b
a
Th
a
n
ka
p
p
a
n
Mer
cy
)
1575
r
etin
al
b
o
r
d
e
r
s
b
y
s
ee
k
in
g
t
h
e
s
h
o
r
test
-
p
ath
b
etwe
en
two
e
n
d
-
n
o
d
es,
u
s
in
g
th
e
g
r
ap
h
s
ea
r
ch
alg
o
r
ith
m
with
Dijk
s
tr
a
'
s
alg
o
r
ith
m
.
T
h
e
s
ea
r
ch
r
eg
io
n
'
s
r
estrictio
n
an
d
d
o
wn
s
am
p
lin
g
h
elp
ed
to
lo
wer
t
h
e
tim
e
co
m
p
lex
ity
.
SVM
with
ch
ar
ac
ter
is
t
ics
d
ep
en
d
en
t
o
n
p
ictu
r
e
in
ten
s
ities
a
n
d
g
r
ad
ien
ts
wer
e
u
tili
ze
d
b
y
Ver
m
ee
r
et
a
l.
[
2
8
]
to
id
en
tify
f
iv
e
r
etin
al
in
ter
f
a
ce
s
.
I
n
o
r
d
er
to
d
eter
m
in
e
th
e
in
itial
s
u
r
f
ac
es,
Srin
iv
asan
et
a
l.
[
2
9
]
co
u
p
le
d
SVM
with
th
e
av
er
ag
e
-
in
te
n
s
ity
o
f
e
v
er
y
r
o
w.
Sin
ce
th
e
d
if
f
er
en
ce
am
o
n
g
ad
jace
n
t
s
u
r
f
a
ce
s
in
n
o
r
m
al
e
y
es
was
s
ig
n
if
ican
tly
lar
g
er
th
an
i
n
OC
T
p
ictu
r
es
o
f
d
am
ag
ed
r
etin
as,
th
e
ch
ar
ac
ter
is
tics
wer
e
p
r
im
ar
ily
cr
ea
ted
f
o
r
ed
g
e
class
if
icatio
n
o
f
o
r
d
i
n
ar
y
ey
es.
T
h
e
s
eg
m
en
tatio
n
o
f
r
etin
al
an
o
m
alies
was
th
e
f
o
cu
s
o
f
m
an
y
m
o
r
e
tech
n
iq
u
es
[
3
0
]
.
An
au
to
m
ated
tech
n
iq
u
e
f
o
r
s
y
m
p
to
m
atic
-
ex
u
d
ate
-
ass
o
ciate
d
d
er
an
g
em
en
ts
(
SEAD
)
f
o
o
tm
ar
k
d
e
d
u
ctio
n
was
d
escr
ib
ed
b
y
Qu
ellec
et
a
l.
[
3
1
]
.
T
h
e
t
o
p
a
n
d
b
o
tto
m
lay
er
s
o
f
t
h
e
OC
T
wer
e
s
eg
m
en
ted
u
s
in
g
a
g
r
ap
h
c
u
t
m
eth
o
d
,
th
e
s
u
b
r
etin
al
f
lu
i
d
an
d
s
u
b
-
R
PE
f
lu
id
am
o
n
g
th
e
s
u
r
f
ac
es
wer
e
s
eg
m
en
ted
u
s
in
g
a
s
p
lit
B
r
eg
m
an
-
b
ased
s
eg
m
en
tatio
n
m
et
h
o
d
[
3
2
]
.
Fin
ally
,
th
e
ac
tu
al
f
lu
id
r
eg
io
n
s
wer
e
s
eg
m
en
ted
u
s
in
g
a
r
a
n
d
o
m
f
o
r
est
class
if
ier
th
at
h
ad
b
ee
n
d
ev
elo
p
e
d
.
Diab
etes
-
r
elate
d
m
ac
u
lar
ed
em
a
was
s
eg
m
en
ted
b
y
W
an
g
et
a
l.
[
3
3
]
u
s
in
g
a
f
u
zz
y
-
le
v
el
s
et
-
d
ep
en
d
en
t seg
m
en
t te
ch
n
iq
u
e.
3.
M
E
T
H
O
D
PED
an
d
n
o
r
m
al
im
ag
es
m
ay
b
e
f
o
u
n
d
i
n
th
e
OC
T
im
ag
e
d
atab
ase,
wh
ich
will
m
ak
e
cl
ass
if
y
in
g
it
m
o
r
e
ef
f
ec
tiv
e
an
d
p
r
ec
is
e.
OC
T
s
ca
n
s
f
r
o
m
n
o
r
m
al
r
etin
a
an
d
PED
ca
n
h
av
e
d
if
f
er
e
n
t
lev
els
o
f
illn
ess
;
a
p
r
o
f
ess
io
n
al
o
r
ML
e
x
p
er
t
is
n
ee
d
ed
to
h
elp
th
e
p
r
ac
titi
o
n
er
class
if
y
th
e
f
i
n
d
in
g
s
.
Fig
u
r
e
5
d
ep
icts
th
e
s
u
g
g
ested
m
eth
o
d
o
lo
g
y
.
Pre
p
r
o
ce
s
s
in
g
,
d
ataset
,
tr
ain
in
g
an
d
class
if
icatio
n
m
ak
e
u
p
t
h
e
t
ec
h
n
iq
u
e
p
ar
t.
T
h
e
co
lo
r
im
ag
e
was
ch
an
g
e
d
to
a
g
r
ay
s
ca
le
v
er
s
io
n
f
o
r
an
aly
s
is
,
an
d
th
e
p
ix
el
v
al
u
es
wer
e
ex
p
an
d
ed
to
co
v
er
th
e
wh
o
le
r
an
g
e
b
etwe
en
0
an
d
1
.
T
h
e
cr
ea
tio
n
o
f
a
m
ac
u
lar
h
o
l
e
is
th
en
d
eter
m
in
e
d
u
s
in
g
th
e
s
u
g
g
ested
m
eth
o
d
s
af
ter
p
r
o
ce
s
s
in
g
th
is
n
o
r
m
aliz
ed
im
ag
e.
T
h
e
n
o
r
m
alize
d
im
ag
e
is
ex
p
o
s
ed
t
o
th
e
m
e
d
ian
f
ilter
to
d
ec
r
ea
s
e
n
o
is
e,
p
r
o
d
u
cin
g
th
e
f
ilter
in
g
im
ag
e
[
3
4
]
.
I
n
th
e
p
r
ep
r
o
ce
s
s
in
g
we
u
s
ed
two
s
tep
s
.
T
h
e
f
ir
s
t
s
tep
d
en
o
is
in
g
an
d
th
e
s
ec
o
n
d
s
tep
R
PE
lay
er
s
eg
m
en
tatio
n
.
Fig
u
r
e
5
.
Pro
p
o
s
ed
m
et
h
o
d
o
lo
g
y
3
.
1
.
Deno
is
ing
Sp
ec
k
le
n
o
is
e
is
c
o
m
m
o
n
ly
o
b
s
er
v
ed
in
OC
T
im
a
g
es,
d
eg
r
ad
in
g
th
eir
q
u
ality
s
ig
n
if
i
ca
n
tly
.
T
o
m
itig
ate
th
is
is
s
u
e,
we
em
p
lo
y
ed
a
W
ien
er
f
ilter
in
g
tech
n
iq
u
e
to
r
em
o
v
e
s
p
ec
k
le
n
o
i
s
e
an
d
co
n
v
er
t
th
e
o
r
ig
in
al
im
ag
e
in
to
g
r
ay
s
ca
le.
Fig
u
r
e
6
s
h
o
ws th
e
o
u
tco
m
e
o
f
th
e
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
with
o
r
ig
in
al
im
ag
e
i
n
Fig
u
r
e
6
(
a)
,
g
r
ay
s
ca
le
im
ag
e
i
n
Fig
u
r
e
6
(
b
)
an
d
d
en
o
is
ed
im
ag
e
in
Fig
u
r
e
6
(
c
)
.
W
e
an
aly
z
ed
th
e
v
ar
io
u
s
ty
p
es
o
f
f
ilter
in
g
tech
n
iq
u
e
s
u
ch
as
m
ea
n
,
m
ed
ian
,
g
au
s
s
ian
,
b
ila
ter
al
an
d
wien
er
.
B
ased
o
n
th
e
m
etr
ices
s
u
ch
a
s
MSE
,
PS
N
R
an
d
PS
R
wien
er
f
ilter
is
th
e
ac
cu
r
ate
d
ec
r
ea
s
e
t
h
e
s
p
ec
k
le
n
o
is
e
[
3
5
]
.
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
.
2
,
Ap
r
il
20
25
:
1
5
7
2
-
1
5
8
3
1576
3
.
2
.
RP
E
la
y
er
s
eg
m
ent
a
t
i
o
n
I
n
th
e
s
ec
o
n
d
s
tep
o
f
p
r
ep
r
o
ce
s
s
in
g
,
we
p
er
f
o
r
m
R
PE
-
lay
er
s
eg
m
en
tatio
n
.
Fo
r
th
is
s
eg
m
e
n
tatio
n
,
we
u
tili
ze
th
e
th
r
esh
o
ld
in
g
m
eth
o
d
.
T
h
e
R
PE
lay
er
ex
h
ib
its
u
n
i
f
o
r
m
b
r
i
g
h
tn
ess
an
d
s
m
o
o
th
n
ess
,
allo
win
g
u
s
t
o
an
aly
ze
ea
ch
p
ix
el
v
alu
e
alo
n
g
th
e
co
lu
m
n
.
Fig
u
r
e
7
s
h
o
w
s
th
e
o
u
t
p
u
t
o
f
t
h
e
R
PE
lay
er
s
eg
m
en
tatio
n
with
R
PE
lay
er
o
n
h
ea
lth
y
g
r
a
y
s
ca
le
im
ag
e
in
Fig
u
r
e
7
(
a)
,
R
PE
o
v
er
lay
o
n
th
e
o
r
i
g
in
al
h
ea
lth
y
im
ag
e
in
Fig
u
r
e
7
(
b
)
,
R
PE
-
lay
er
with
PED
o
n
a
g
r
ay
s
ca
le
im
ag
e
in
Fig
u
r
e
7
(
c)
an
d
R
PE
lay
er
o
n
PED
-
im
ag
e
in
Fig
r
u
e
7
(
d
)
.
W
e
an
al
y
ze
d
th
e
R
PE
lay
er
s
eg
m
en
tatio
n
with
2
0
im
ag
es
o
f
h
ea
lth
y
a
n
d
2
0
im
ag
es
o
f
PED
im
ag
es
an
d
g
o
t
h
ea
lth
y
m
ac
u
la
f
u
lly
d
etec
ted
co
r
r
ec
tly
a
n
d
PED
1
9
im
ag
es
d
etec
ted
co
r
r
ec
tly
.
On
e
k
e
y
ap
p
r
o
ac
h
in
im
ag
e
s
eg
m
en
tati
o
n
is
th
r
esh
o
ld
in
g
.
T
h
is
m
eth
o
d
m
ay
b
e
d
escr
ib
ed
as
(
1
)
,
=
[
,
,
(
,
)
,
(
,
)
]
(
1
)
(
a)
(
b
)
(
c)
Fig
u
r
e
6
.
Pre
p
r
o
ce
s
s
in
g
:
(
a)
o
r
ig
in
al
im
ag
e,
(
b
)
g
r
ay
-
s
ca
le
i
m
ag
e,
an
d
(
c)
d
en
o
is
ed
im
a
g
e
(
wien
er
f
ilter
)
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
7
.
L
a
y
er
s
eg
m
e
n
tatio
n
:
(
a)
R
PE
lay
er
o
n
h
ea
lth
y
g
r
ay
s
ca
le
im
ag
e
,
(
b
)
R
PE
o
v
e
r
lay
o
n
th
e
o
r
ig
in
al
h
ea
lth
y
im
ag
e
,
(
c)
R
PE
-
lay
er
with
PED
o
n
a
g
r
a
y
s
ca
le
im
ag
e
(
d
)
R
PE
lay
er
o
n
PED
-
im
ag
e
3
.
3
.
E
x
t
r
a
ct
t
he
f
e
a
t
ures
T
h
e
p
r
o
ce
s
s
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
is
cr
u
cial
f
o
r
r
ec
o
g
n
izin
g
th
e
m
ac
u
lar
-
h
o
le
in
th
e
O
C
T
im
ag
e.
Featu
r
e
ex
tr
ac
tio
n
p
r
o
ce
s
s
in
v
o
lv
es
two
p
ar
ts
.
T
h
e
lis
t
o
f
b
o
r
d
er
p
ix
els
f
o
r
ea
ch
co
l
u
m
n
is
o
n
ly
u
s
ed
to
ch
o
o
s
e
o
n
e
p
i
x
el
f
r
o
m
ea
c
h
c
o
lu
m
n
i
n
th
e
f
ir
s
t
s
tep
.
T
h
e
2
nd
s
tep
id
en
tifie
s
th
e
2
lar
g
est
y
-
ax
is
d
if
f
er
e
n
ce
s
b
etwe
en
2
s
u
cc
ess
iv
e
p
el
[
3
6
]
–
[
3
8
]
.
T
h
e
ch
ar
ac
ter
is
tics
o
f
i
m
ag
es
h
av
e
l
o
n
g
b
ee
n
m
an
u
a
lly
r
etr
iev
ed
u
s
in
g
v
ar
io
u
s
f
ea
tu
r
e
-
e
n
g
in
ee
r
i
n
g
te
ch
n
iq
u
es in
class
ic
m
ac
h
in
e
le
ar
n
in
g
m
eth
o
d
s
.
T
h
e
ar
t o
f
ex
t
r
ac
t sig
n
if
ican
t a
n
d
v
alu
ab
le
p
atter
n
f
r
o
m
d
ata
th
at
n
ec
k
ed
ey
es
ca
n
n
o
t
s
ee
is
k
n
o
wn
as
f
ea
tu
r
e
en
g
in
ee
r
in
g
.
Ma
ch
in
e
lear
n
i
n
g
m
o
d
els
ca
n
id
en
tif
y
b
etwe
en
class
es
m
o
r
e
ea
s
ily
b
y
ex
tr
ac
tin
g
f
ea
tu
r
es.
B
ec
au
s
e
it
r
ed
u
ce
s
th
e
am
o
u
n
t
o
f
lab
els
th
at
m
u
s
t
b
e
co
n
s
id
er
ed
f
o
r
p
r
ec
is
e
class
if
icatio
n
,
th
is
ch
ar
ac
ter
is
tic
is
b
en
ef
icial
f
o
r
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
.
T
h
er
e
f
o
r
e,
f
ea
t
u
r
e
en
g
in
ee
r
in
g
is
th
e
m
o
s
t
cr
u
cial
ab
ilit
y
f
o
r
ac
h
iev
in
g
g
o
o
d
r
esu
lts
f
o
r
th
e
m
ajo
r
ity
o
f
p
r
e
d
ictio
n
task
s
[
3
9
]
.
T
r
ad
itio
n
al
im
ag
e
class
if
icatio
n
an
d
p
r
e
d
ictio
n
tech
n
iq
u
e
s
in
clu
d
e
m
ac
h
in
e
lear
n
in
g
class
if
ier
s
an
d
r
eg
r
ess
o
r
s
lik
e
KNN
,
SVM,
an
d
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN)
.
C
o
lo
r
,
ed
g
es,
f
o
r
m
,
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
N
o
ve
l fe
a
tu
r
es e
xtra
ctio
n
:
p
ig
men
t e
p
ith
elia
l d
eta
ch
me
n
t
d
e
tectio
n
u
s
in
g
…
(
S
h
ee
b
a
Th
a
n
ka
p
p
a
n
Mer
cy
)
1577
tex
tu
r
e,
lo
ca
l
b
in
a
r
y
p
atter
n
s
(
L
B
P),
an
d
h
is
to
g
r
am
s
o
f
th
e
o
r
ien
ted
g
r
ad
ien
ts
(
HOG)
a
r
e
f
u
r
th
er
h
an
d
-
d
r
awn
p
ictu
r
e
ch
ar
ac
ter
is
tics
.
Af
ter
th
e
f
ea
tu
r
es
h
av
e
b
ee
n
r
ec
o
v
er
ed
,
a
class
if
ier
is
tr
ain
ed
to
ca
r
r
y
o
u
t
th
e
class
if
icatio
n
,
d
etec
tio
n
,
o
r
p
r
ed
ictio
n
jo
b
f
o
r
th
e
im
ag
es
[
2
]
.
I
n
th
is
w
o
r
k
to
d
etec
ted
th
e
PED,
we
a
n
aly
ze
d
m
ac
h
in
e
lear
n
in
g
class
if
icatio
n
s
ar
e
d
ec
is
io
n
tr
ee
(
DT
)
,
lo
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
,
n
aiv
e
B
ay
es
(
N
B
)
,
ANN,
R
F,
SVM
an
d
KNN.
T
o
p
r
e
d
icted
th
e
PED
we
u
s
ed
f
o
u
r
f
ea
tu
r
es
ar
e
ex
tr
ac
ted
s
u
ch
a
s
Ma
x_
Left_
Heig
h
t
(
M_
L
_
H
)
,
N
u
mb
er_
o
f_
Left_
Do
w
n
_
P
o
in
ts
(
N_
L
_
D_
P
)
,
N
o
_
o
f_
R
i
g
h
t_
Do
w
n
_
P
o
in
ts
(
N_
R
_
D_
P
)
a
n
d
Ma
x_
R
ig
h
t_
Heig
h
t
(
M_
R
_
H
)
.
T
h
ese
f
ea
tu
r
es
ar
e
s
ig
n
if
ic
an
tly
p
r
ed
icted
t
h
e
PED
d
is
ea
s
e
an
d
Fig
u
r
e
8
d
ep
icts
th
e
o
u
tco
m
e
o
f
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
with
PED
im
a
g
e
with
R
PE
-
lay
er
in
Fig
u
r
e
8
(
a)
,
s
ep
ar
ated
R
PE
-
lay
er
in
Fig
u
r
e
8
(
b
)
an
d
ex
tr
ac
ted
f
o
u
r
f
ea
tu
r
es in
Fig
u
r
e
8
(
c
)
.
(
a)
(
b
)
(
c)
Fig
u
r
e
8
.
Featu
r
e
ex
tr
ac
tio
n
:
(
a)
PED
im
ag
e
with
R
PE
-
lay
er
,
(
b
)
s
ep
ar
ate
d
R
PE
-
lay
er
,
an
d
(
c)
ex
tr
ac
ted
f
o
u
r
f
ea
tu
r
es
Her
e,
we
u
s
ed
(
2
)
-
(
4
)
to
f
i
n
d
o
u
t th
e
to
p
-
la
y
er
,
b
o
tto
m
-
la
y
er
an
d
m
ax
-
h
eig
h
t:
=
min
(
)
(
2
)
=
ma
x
(
)
(
3
)
ℎ
=
−
(
4
)
wh
er
e
is
th
e
R
PE
lay
er
's
y
ax
is
v
alu
e
f
o
r
ea
c
h
x
a
x
is
v
alu
e,
wh
er
e
x
e
x
ten
d
s
f
r
o
m
ze
r
o
to
th
e
wid
th
o
f
th
e
im
ag
e.
B
ec
au
s
e
th
e
to
p
-
lef
t c
o
r
n
er
o
f
th
e
im
ag
e
is
at
(
0
,
0
)
,
t
h
e
‘
m
in
’
m
eth
o
d
is
u
s
ed
to
g
et
th
e
s
m
allest y
-
ax
is
v
alu
e
o
f
t
h
e
R
PE
lay
er
,
in
o
r
d
er
to
o
b
tain
th
e
to
p
o
f
t
h
e
la
y
er
.
T
h
e
b
o
tto
m
m
o
s
t
lay
er
is
id
en
tifie
d
u
s
in
g
th
e
‘
m
ax
’
f
u
n
ctio
n
.
T
h
e
to
p
-
lef
t
p
o
in
t
is
th
en
d
eter
m
in
e
d
b
y
r
ep
ea
tin
g
th
e
li
n
e
f
r
o
m
ze
r
o
t
o
th
e
lin
e'
s
h
ig
h
est
p
o
in
t.
I
te
r
atin
g
f
r
o
m
th
e
esti
m
ated
to
p
-
le
f
t
p
o
i
n
t
to
t
h
e
li
n
e'
s
h
ig
h
est
p
o
in
t
y
ield
s
th
e
to
p
-
r
ig
h
t
p
o
i
n
t.
T
h
e
to
p
-
lef
t
p
o
in
t
is
th
en
u
s
ed
t
o
ex
tr
ac
t
2
0
p
o
i
n
ts
to
b
u
ild
a
b
o
u
n
d
in
g
r
ec
ta
n
g
le.
T
h
e
R
PE
lay
er
h
eig
h
t
g
o
v
er
n
s
th
e
wid
th
o
f
t
h
e
r
ec
tan
g
le.
=
ℎ
(
5
)
I
n
th
is
s
tu
d
y
,
we
ex
t
r
ac
ted
2
0
p
o
in
ts
f
r
o
m
t
h
e
to
p
-
lef
t
co
r
n
er
a
n
d
2
0
p
o
in
ts
f
r
o
m
th
e
to
p
-
r
ig
h
t
co
r
n
er
to
g
o
v
er
n
s
th
e
cu
r
v
e.
T
h
e
in
ter
v
a
l is ca
lcu
lated
f
o
r
th
e
2
0
p
o
in
t
s
ex
tr
ac
ted
f
r
o
m
th
e
to
p
-
lef
t c
o
r
n
er
as
(
6
)
:
=
20
(
6
)
T
h
e
lin
e
h
ei
g
h
t
o
f
ea
ch
p
o
in
t
is
th
en
o
b
tain
e
d
b
y
iter
atin
g
th
e
lin
e
p
o
in
ts
2
0
tim
es
b
ac
k
war
d
f
r
o
m
th
e
to
p
-
le
f
t
p
o
in
t:
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
.
2
,
Ap
r
il
20
25
:
1
5
7
2
-
1
5
8
3
1578
=
ℎ
−
(
(
−
∗
)
)
(
7
)
wh
er
e
d
if
f
er
s
f
r
o
m
1
to
2
0
.
T
h
e
s
u
b
s
eq
u
en
t e
q
u
atio
n
ex
tr
a
cted
th
e
to
p
-
r
ig
h
t
2
0
p
o
in
ts
:
=
ℎ
−
(
(
ℎ
+
∗
)
)
(
8
)
wh
er
e
v
ar
ies f
r
o
m
1
to
2
0
.
T
h
e
eq
u
atio
n
s
(
9
)
an
d
(
1
0
)
ar
e
u
s
ed
to
ex
tr
ac
t
th
e
m
ax
im
u
m
lef
t
h
eig
h
t
(
M_
L
_
H)
an
d
m
ax
im
u
m
r
ig
h
t h
eig
h
t
(
M_
R
_
H)
f
ea
t
u
r
es:
_
_
=
(
)
(
9
)
_
_
=
(
)
(
1
0
)
B
y
co
m
p
ar
in
g
t
h
e
s
u
b
s
eq
u
en
t
lef
t
-
p
o
in
ts
,
th
e
n
u
m
b
er
o
f
lef
t
d
o
wn
p
o
i
n
ts
(
N_
L
_
D_
P)
ch
ar
ac
ter
is
tic
is
th
en
co
m
p
u
ted
.
I
n
cr
ea
s
e
th
e
co
u
n
t
er
N_
L
_
D
_
P
if
a
lef
t
-
p
o
in
t
v
a
lu
e
is
s
m
aller
th
a
n
t
h
e
f
o
llo
win
g
lef
t
-
p
o
in
t
v
alu
e
.
B
y
co
m
p
ar
in
g
th
e
r
ig
h
t
-
p
o
i
n
ts
an
d
in
cr
ea
s
in
g
th
e
co
u
n
ter
,
th
e
n
u
m
b
er
o
f
r
ig
h
t
d
o
wn
p
o
in
ts
(
N_
R
_
D_
P)
f
ea
tu
r
e
is
d
eter
m
in
ed
.
T
h
e
a
lg
o
r
ith
m
to
d
etec
t
th
e
f
o
u
r
d
is
tin
ct
f
ea
tu
r
es
to
d
etec
t
P
E
D
is
g
iv
en
in
th
e
Alg
o
r
ith
m
1
.
_
_
_
=
_
_
_
+
1
(
<
(
+
1
)
(
1
1
)
_
_
_
=
_
_
_
+
1
(
<
(
+
1
)
)
(
1
2
)
Alg
o
r
ith
m
1
.
E
x
tr
ac
t_
Featu
r
es
(
)
I
N
P
U
T:
L
-
A
n
a
r
r
a
y
c
o
n
t
a
i
n
s
y
v
a
l
u
e
s
o
f
e
a
c
h
c
o
l
u
m
n
p
i
x
e
l
o
f
t
h
e
l
a
y
e
r
O
U
TPU
T:
maxLeftHeight
–
M
a
x
i
m
u
m
l
e
f
t
h
e
i
g
h
t
maxRightHeight
–
M
a
x
i
m
u
m rig
h
t
h
e
i
g
h
t
noOfLeftDownPoints
–
N
u
mb
e
r
o
f
l
e
f
t
d
o
w
n
p
o
i
n
t
s
noOfRightDownPoints
–
N
u
mb
e
r
o
f
r
i
g
h
t
d
o
w
n
p
o
i
n
t
s
S
t
e
p
-
1
:
=
(
)
S
t
e
p
-
2
:
=
(
)
S
t
e
p
-
3
:
ℎ
=
−
;
S
t
e
p
-
4
:
ℎ
=
−
;
S
t
e
p
-
5
:
S
t
e
p
-
6
=
1
ℎ
(
)
S
t
e
p
-
6
:
(
ℎ
=
ℎ
)
S
t
e
p
-
7
S
t
e
p
-
7
:
=
S
t
e
p
-
8
:
=
ℎ
S
t
e
p
-
9
:
=
20
S
t
e
p
-
1
0
:
=
/
;
S
t
e
p
-
1
1
:
(
1
)
=
ℎ
;
S
t
e
p
-
1
2
:
S
t
e
p
-
1
3
S
t
e
p
-
1
4
=
1
S
t
e
p
-
1
3
:
=
−
_
∗
;
S
t
e
p
-
1
4
:
(
+
1
)
=
ℎ
(
)
;
S
t
e
p
-
1
5
:
S
t
e
p
-
1
6
=
(
ℎ
)
S
t
e
p
-
1
6
:
(
ℎ
(
)
!
=
ℎ
)
S
t
e
p
-
17
S
t
e
p
-
1
7
:
ℎ
=
;
S
t
e
p
-
1
8
:
(
1
)
=
ℎ
;
S
t
e
p
-
1
9
:
S
t
e
p
-
2
0
S
t
e
p
-
2
1
=
1
S
t
e
p
-
2
0
:
=
ℎ
+
∗
S
t
e
p
-
2
1
:
ℎ
(
+
1
)
=
ℎ
(
)
S
t
e
p
-
2
2
:
ℎ
=
(
ℎ
)
;
S
t
e
p
-
2
3
:
ℎ
ℎ
=
(
ℎ
ℎ
)
;
S
t
e
p
-
2
4
:
=
0
;
S
t
e
p
-
2
5
:
ℎ
=
0
;
S
t
e
p
-
2
6
:
S
t
e
p
-
2
7
S
t
e
p
-
3
0
=
2
S
t
e
p
-
2
7
:
(
(
ℎ
−
ℎ
−
1
)
>
0
)
S
t
e
p
-
2
8
:
=
+
1
;
S
t
e
p
-
2
9
:
(
(
ℎ
−
ℎ
−
1
)
>
0
)
S
t
e
p
-
3
0
:
ℎ
=
ℎ
+
1
;
S
t
e
p
-
3
1
:
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
N
o
ve
l fe
a
tu
r
es e
xtra
ctio
n
:
p
ig
men
t e
p
ith
elia
l d
eta
ch
me
n
t
d
e
tectio
n
u
s
in
g
…
(
S
h
ee
b
a
Th
a
n
ka
p
p
a
n
Mer
cy
)
1579
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
o
r
ig
in
al
OC
T
im
ag
es
wer
e
co
n
v
er
te
d
in
to
g
r
a
y
s
ca
le,
an
d
th
en
we
ap
p
lied
t
h
e
W
ien
er
f
ilter
in
g
tech
n
iq
u
e
to
e
f
f
ec
tiv
ely
r
em
o
v
e
s
p
ec
k
le
n
o
is
e,
t
h
en
e
x
tr
ac
t
ed
th
e
R
PE
lay
er
u
s
ed
to
th
r
e
s
h
o
ld
m
eth
o
d
,
wh
ile
PED
f
o
u
n
d
i
n
th
e
R
PE
lay
er
,
th
en
to
d
etec
ted
th
e
PED.
W
e
ex
tr
ac
ted
f
o
u
r
f
ea
tu
r
es
wer
e
to
id
en
tifie
d
PED
with
m
ac
h
in
e
lear
n
in
g
class
if
ier
s
.
T
o
al
2
0
0
im
a
g
es
am
o
n
g
1
0
0
im
ag
es
o
f
n
o
r
m
al
r
etin
a,
1
0
0
im
ag
es
o
f
PED,
a
lo
ca
l
d
ataset.
R
etin
al
OC
T
im
ag
es
with
n
o
r
m
al
a
n
d
PED
r
etin
as
ar
e
in
clu
d
ed
in
t
h
e
te
s
tin
g
d
ataset
alo
n
g
with
a
tr
ain
in
g
d
ataset.
I
n
th
is
wo
r
k
o
r
d
in
ar
y
r
etin
al
im
ag
es
ar
e
ass
ig
n
ed
a
v
alu
e
o
f
0
,
wh
er
ea
s
PED
r
etin
al
im
ag
e
is
ass
ig
n
ed
a
v
alu
e
o
f
1
.
B
ased
o
n
th
e
T
ab
le
1
,
an
al
y
ze
d
th
e
tr
u
e
-
p
o
s
itiv
e
(T
-
P),
f
alse
-
n
eg
ativ
e
(F
-
N)
,
tr
u
e
-
n
eg
ativ
e
(T
-
N)
a
n
d
f
alse
-
p
o
s
itiv
e
(F
-
P)
p
ar
am
eter
s
a
n
d
Fig
u
r
e
9
d
ep
icts
th
e
g
r
ap
h
ica
l
r
ep
r
esen
tatio
n
o
f
th
ese
p
ar
am
eter
s
.
B
ased
o
n
th
e
T
ab
le
2
,
to
c
alcu
late
th
e
f
o
llo
win
g
m
etr
ic
s
ar
e
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
s
p
ec
if
icity
,
p
r
ec
is
io
n
an
d
F1
-
s
co
r
e
an
d
Fig
u
r
e
1
0
s
h
o
ws
th
e
g
r
a
p
h
ical
r
ep
r
esen
tatio
n
o
f
th
ese
m
etr
i
ce
s
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
m
a
y
b
e
u
s
ed
to
d
eter
m
in
e
th
e
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
s
p
ec
if
icity
an
d
p
r
ec
is
io
n
.
Fig
u
r
e
1
1
d
ep
icts
th
e
co
n
f
u
s
io
n
m
atr
i
x
f
o
r
th
e
SVM
m
o
d
el.
B
ased
o
n
th
e
Fig
u
r
e
1
0
,
th
e
r
an
d
o
m
f
o
r
est (
R
F)
m
o
d
el
is
p
er
f
o
r
m
ed
th
e
b
est
class
if
icatio
n
m
o
d
el
c
o
m
p
ar
in
g
th
e
o
th
er
class
if
ier
s
.
R
F
was
p
r
o
v
i
d
ed
f
o
r
1
0
0
%
ac
cu
r
ac
y
,
1
0
0
%
s
en
s
itiv
ity
,
1
0
0
%
s
p
ec
if
icity
,
1
0
0
%
p
r
ec
is
io
n
,
an
d
1
0
0
%
F1
-
s
co
r
e.
T
h
e
ar
ea
u
n
d
er
th
e
R
OC
cu
r
v
e
(
AUC)
o
f
r
ec
eiv
er
o
p
e
r
atin
g
ch
a
r
ac
ter
is
tics
(
R
O
C
)
cu
r
v
e
will
b
e
u
s
ed
to
lin
k
th
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
ML
class
if
ier
with
b
in
ar
y
class
if
icatio
n
.
T
h
e
R
F
-
ML
m
o
d
el'
s
AU
C
was
1
.
0
.
T
h
e
R
F
-
ML
class
if
ier
m
o
d
el
s
u
g
g
ested
th
at
g
lo
b
al
ac
h
iev
em
en
t
is
s
u
p
er
io
r
in
class
if
y
in
g
am
o
n
g
PED
a
n
d
o
r
d
in
ar
y
r
etin
a
o
f
OC
T
-
im
ag
es
b
ec
au
s
e
t
h
e
AUC
v
alu
e
is
to
1
(
h
ig
h
)
,
as sh
o
wn
i
n
Fig
u
r
e
1
2
.
T
ab
le
1
.
R
esu
lt o
f
th
e
m
ac
h
in
e
lear
n
in
g
class
if
icatio
n
s
an
d
A
NN
C
l
a
s
si
f
i
e
r
s
T
-
P
F
-
N
T
-
N
F
-
P
S
V
M
20
0
18
2
K
N
N
18
2
18
2
LR
20
0
19
1
DT
20
0
18
2
RF
20
0
20
0
NB
20
0
18
2
B
P
N
N
20
0
18
2
N
o
t
e
:
B
P
N
N
-
b
a
c
k
p
r
o
p
a
g
a
t
i
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
Fig
u
r
e
9
.
An
al
y
s
is
o
f
class
if
ie
r
s
u
s
in
g
T
-
P,
F
-
N,
T
-
N
a
n
d
F
-
P
T
ab
le
2
.
An
aly
s
is
r
ep
o
r
t o
f
m
a
ch
in
e
lear
n
in
g
class
if
ier
s
an
d
B
PNN
ML
-
c
l
a
ss
i
f
i
e
r
s
A
c
c
u
r
a
c
y
S
e
n
s
i
t
i
v
i
t
y
S
p
e
c
i
f
i
c
i
t
y
P
r
e
c
i
s
i
o
n
F1
-
sc
o
r
e
S
V
M
0
.
9
5
1
.
0
0
0
.
9
0
.
9
0
9
1
0
.
9
5
2
4
K
N
N
0
.
9
0
.
9
0
0
.
9
0
.
9
0
.
9
LR
0
.
9
7
5
1
.
0
0
0
.
9
5
0
.
9
5
2
4
0
.
9
7
5
6
DT
0
.
9
5
1
.
0
0
0
.
9
0
.
9
0
9
1
0
.
9
5
2
4
RF
1
.
0
0
1
.
0
0
1
.
0
0
1
.
0
0
1
.
0
0
NB
0
.
9
5
1
.
0
0
0
.
9
0
0
.
9
1
0
.
9
5
2
4
B
P
N
N
0
.
9
5
1
.
0
0
0
.
9
0
0
.
9
1
0
.
9
5
2
4
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
.
2
,
Ap
r
il
20
25
:
1
5
7
2
-
1
5
8
3
1580
Fig
u
r
e
1
0
.
An
aly
s
is
o
f
m
etr
ics o
f
m
ac
h
i
n
e
lear
n
in
g
class
if
ier
s
an
d
B
PNN
Fig
u
r
e
1
1
.
C
o
n
f
u
s
io
n
m
atr
ix
o
f
SVM
Fig
u
r
e
1
2
.
R
OC
cu
r
v
e
f
o
r
SV
M
an
d
AUC f
o
r
0
.
9
8
7
8
7
8
7
8
7
8
7
8
7
8
7
9
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
N
o
ve
l fe
a
tu
r
es e
xtra
ctio
n
:
p
ig
men
t e
p
ith
elia
l d
eta
ch
me
n
t
d
e
tectio
n
u
s
in
g
…
(
S
h
ee
b
a
Th
a
n
ka
p
p
a
n
Mer
cy
)
1581
5.
CO
NCLU
SI
O
N
T
h
e
p
r
im
ar
y
c
o
n
tr
ib
u
tio
n
o
f
t
h
is
wo
r
k
is
,
th
e
ex
tr
ac
tio
n
o
f
f
ea
tu
r
es
f
r
o
m
OC
T
PED
im
ag
e
s
.
I
n
OC
T
,
wh
ich
co
n
s
is
ts
o
f
PED
an
d
n
o
r
m
al
r
etin
a,
m
ay
b
e
d
etec
ted
PED
u
s
in
g
m
ac
h
i
n
e
lear
n
in
g
b
ased
class
if
ier
m
o
d
els
an
d
b
ac
k
p
r
o
p
ag
atio
n
n
e
u
r
al
n
etwo
r
k
.
T
o
ass
ess
class
if
icatio
n
p
ar
am
eter
s
i
n
clu
d
in
g
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
s
p
ec
if
icity
,
p
r
ec
is
i
o
n
,
an
d
F1
-
s
co
r
e
,
v
ar
io
u
s
m
ac
h
in
e
lear
n
in
g
class
if
ier
m
o
d
el
s
wer
e
u
tili
ze
d
.
T
h
e
SVM
-
ac
cu
r
ac
y
,
SVM
-
s
en
s
itiv
ity
,
SVM
-
s
p
ec
if
icity
,
SVM
-
p
r
ec
is
io
n
an
d
SVM
-
F
1
-
s
co
r
e
o
f
SVM
is
9
5
%,
1
0
0
%,
9
0
%,
9
0
%
an
d
9
5
%,
KNN
is
9
0
%,
9
0
%,
9
0
%,
9
0
%
a
n
d
9
0
%,
L
R
is
9
7
%,
1
0
0
%,
9
5
%,
9
5
%
an
d
9
7
%,
DT
is
9
5
%,
1
0
0
%,
9
0
%,
9
0
% a
n
d
9
5
%,
R
F is
1
0
0
%,
1
0
0
%,
1
0
0
%,
1
0
0
% a
n
d
1
0
0
%,
NB
is
9
5
%,
1
0
0
%,
9
0
%,
9
1
%
an
d
9
5
%
an
d
ANN
is
9
5
%,
1
0
0
%,
9
0
%,
9
1
%
an
d
9
5
%.
T
h
e
L
R
R
OC
cu
r
v
e
h
ad
an
AUC
o
f
0
.
9
9
3
9
,
wh
ile
th
e
DT
was
0
.
9
3
3
3
,
th
e
SVM
was
0
.
9
8
7
8
7
8
,
th
e
K
NN's
was
0
.
9
4
8
4
8
4
8
4
8
4
8
4
8
6
,
an
d
th
e
NB
was
0
.
9
9
3
9
3
,
wh
ile
th
e
R
F
was
1
.
0
.
T
h
e
R
F
p
er
f
o
r
m
e
d
as
th
e
b
est
class
if
ier
to
d
etec
ted
th
e
PED
b
ased
o
n
t
h
e
af
o
r
em
en
tio
n
ed
r
esu
lts
f
o
r
th
e
class
if
icat
io
n
o
f
ML
.
I
n
th
e
ca
teg
o
r
izatio
n
o
f
ML
n
o
r
m
al
an
d
PED,
KNN
p
er
f
o
r
m
s
th
e
p
o
o
r
est.
T
h
e
f
i
n
d
in
g
s
o
f
th
e
m
ac
h
in
e
lear
n
i
n
g
-
b
ased
m
o
d
els’
r
an
d
o
m
f
o
r
est
,
d
ec
is
io
n
tr
ee
,
lo
g
is
tic
r
eg
r
ess
io
n
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
,
an
d
n
aiv
e
B
ay
es
d
ev
elo
p
ed
b
y
class
if
y
in
g
m
etr
ics
u
s
in
g
th
e
p
r
o
p
er
tech
n
i
q
u
e
wer
e
ef
f
icie
n
t
an
d
p
r
ac
tical
in
d
etec
ted
n
o
r
m
al
r
etin
al
an
d
PED
in
OC
T
im
ag
es.
Fin
ally
,
in
ter
m
s
o
f
PED
id
en
tific
atio
n
i
n
OC
T
im
ag
es,
th
e
r
an
d
o
m
f
o
r
est
class
if
ier
g
iv
es
b
est
p
er
f
o
r
m
an
ce
am
o
n
g
th
e
o
th
er
s
.
I
n
f
u
tu
r
e
we
will id
en
tify
an
o
t
h
er
d
is
ea
s
e
.
RE
F
E
R
E
NC
E
S
[
1
]
B
.
K
.
Tr
i
w
i
j
o
y
o
,
B
.
S
.
S
a
b
a
r
g
u
n
a
,
W
.
B
u
d
i
h
a
r
t
o
,
a
n
d
E.
A
b
d
u
r
a
c
h
ma
n
,
“
D
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
e
y
e
d
i
s
e
a
se
s
b
a
s
e
d
o
n
c
o
l
o
r
f
u
n
d
u
s
i
ma
g
e
s
,
”
i
n
D
i
a
b
e
t
e
s
a
n
d
F
u
n
d
u
s
O
C
T
,
E
l
se
v
i
e
r
,
2
0
2
0
,
p
p
.
2
5
–
57
,
d
o
i
:
1
0
.
1
0
1
6
/
b
9
7
8
-
0
-
12
-
8
1
7
4
4
0
-
1
.
0
0
0
0
2
-
4.
[
2
]
U
.
F
a
t
e
ma
,
J
.
H
a
sa
n
,
a
n
d
S
.
A
.
H
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
a
n
a
l
y
s
i
s o
f
r
e
t
i
n
a
l
O
C
T
i
m
a
g
e
u
s
i
n
g
h
a
n
d
c
r
a
f
t
e
d
a
n
d
d
e
e
p
l
e
a
r
n
i
ng
f
e
a
t
u
r
e
w
i
t
h
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
,
”
B
.
S
.
t
h
e
s
i
s,
D
e
p
a
r
t
m
e
n
t
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
El
e
c
t
r
o
n
i
c
E
n
g
i
n
e
e
r
i
n
g
,
H
a
j
e
e
M
o
h
a
mm
a
d
D
a
n
e
s
h
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
U
n
i
v
e
r
si
t
y
,
2
0
2
1
.
[
3
]
P
.
K
.
K
a
r
n
a
n
d
W
.
H
.
A
b
d
u
l
l
a
,
“
O
n
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
i
n
c
l
i
n
i
c
a
l
i
n
t
e
r
p
r
e
t
a
t
i
o
n
o
f
r
e
t
i
n
a
l
d
i
se
a
ses
u
s
i
n
g
O
C
T
i
ma
g
e
s,”
Bi
o
e
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
0
,
n
o
.
4
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
b
i
o
e
n
g
i
n
e
e
r
i
n
g
1
0
0
4
0
4
0
7
.
[
4
]
S
.
Za
y
i
t
-
S
o
u
d
r
y
,
I
.
M
o
r
o
z
,
a
n
d
A
.
Lo
e
w
e
n
s
t
e
i
n
,
“
R
e
t
i
n
a
l
p
i
g
m
e
n
t
e
p
i
t
h
e
l
i
a
l
d
e
t
a
c
h
me
n
t
,
”
S
u
rv
e
y
o
f
O
p
h
t
h
a
l
m
o
l
o
g
y
,
v
o
l
.
5
2
,
n
o
.
3
,
p
p
.
2
2
7
–
2
4
3
,
M
a
y
2
0
0
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
s
u
r
v
o
p
h
t
h
a
l
.
2
0
0
7
.
0
2
.
0
0
8
.
[
5
]
S
.
L
a
t
h
a
,
R
.
L
o
g
a
n
a
t
h
a
n
,
a
n
d
K
.
S
h
a
n
mu
g
a
m,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
n
o
r
ma
l
v
s.
a
g
e
-
r
e
l
a
t
e
d
mac
u
l
a
r
d
e
g
e
n
e
r
a
t
i
o
n
O
C
T
i
ma
g
e
s,
”
BO
H
R
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
S
m
a
r
t
C
o
m
p
u
t
i
n
g
a
n
d
I
n
f
o
r
m
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
3
4
–
3
8
,
2
0
2
1
,
d
o
i
:
1
0
.
5
4
6
4
6
/
b
i
j
s
c
i
t
.
2
0
2
1
.
1
6
.
[
6
]
D
.
X
i
a
n
g
e
t
a
l
.
,
“
A
u
t
o
m
a
t
i
c
r
e
t
i
n
a
l
l
a
y
e
r
se
g
m
e
n
t
a
t
i
o
n
o
f
O
C
T
i
ma
g
e
s
w
i
t
h
c
e
n
t
r
a
l
ser
o
u
s
r
e
t
i
n
o
p
a
t
h
y
,
”
I
EEE
J
o
u
r
n
a
l
o
f
Bi
o
m
e
d
i
c
a
l
a
n
d
H
e
a
l
t
h
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
2
3
,
n
o
.
1
,
p
p
.
2
8
3
–
2
9
5
,
J
a
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
j
b
h
i
.
2
0
1
8
.
2
8
0
3
0
6
3
.
[
7
]
D
.
X
i
a
n
g
e
t
a
l
.
,
“
A
u
t
o
m
a
t
i
c
se
g
me
n
t
a
t
i
o
n
o
f
r
e
t
i
n
a
l
l
a
y
e
r
i
n
O
C
T
i
ma
g
e
s
w
i
t
h
c
h
o
r
o
i
d
a
l
n
e
o
v
a
s
c
u
l
a
r
i
z
a
t
i
o
n
,
”
I
E
EE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
I
m
a
g
e
Pro
c
e
ss
i
n
g
,
v
o
l
.
2
7
,
n
o
.
1
2
,
p
p
.
5
8
8
0
–
5
8
9
1
,
D
e
c
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
0
9
/
t
i
p
.
2
0
1
8
.
2
8
6
0
2
5
5
.
[
8
]
Y
.
H
u
a
n
g
a
n
d
J.
H
u
,
“
R
e
si
d
u
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
-
b
a
se
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
m
a
c
u
l
a
r
e
d
e
ma
i
n
O
C
T,
”
i
n
2
0
1
9
I
EEE
3
1
st
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
T
o
o
l
s
w
i
t
h
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
C
T
AI
)
,
N
o
v
.
2
0
1
9
,
p
p
.
7
3
6
–
7
4
3
,
d
o
i
:
1
0
.
1
1
0
9
/
i
c
t
a
i
.
2
0
1
9
.
0
0
1
0
7
.
[
9
]
O
.
L.
C
.
M
e
n
d
e
s
e
t
a
l
.
,
“
A
u
t
o
m
a
t
i
c
se
g
me
n
t
a
t
i
o
n
o
f
mac
u
l
a
r
h
o
l
e
s i
n
o
p
t
i
c
a
l
c
o
h
e
r
e
n
c
e
t
o
m
o
g
r
a
p
h
y
i
ma
g
e
s,
”
I
EE
E
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
9
6
4
8
7
–
9
6
5
0
0
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
a
c
c
e
ss.
2
0
2
1
.
3
0
9
2
3
0
1
.
[
1
0
]
N
.
T.
N
.
N
.
A
z
a
me
n
,
A
.
M
.
A
l
i
,
a
n
d
N
.
A
.
A
.
A
z
i
z
,
“
P
r
e
d
i
c
t
i
o
n
o
f
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
b
a
se
d
o
n
r
i
s
k
f
a
c
t
o
r
s
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms,
”
i
n
2
0
2
3
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
D
a
t
a
S
c
i
e
n
c
e
s
(
Ai
D
AS
)
,
S
e
p
.
2
0
2
3
,
p
p
.
3
0
8
–
312
,
d
o
i
:
1
0
.
1
1
0
9
/
a
i
d
a
s6
0
5
0
1
.
2
0
2
3
.
1
0
2
8
4
6
4
6
.
[
1
1
]
M
.
K
a
r
a
m
p
e
l
a
s
,
P
.
M
a
l
a
m
o
s
,
P
.
P
e
t
r
o
u
,
I
.
G
e
o
r
g
a
l
a
s
,
D
.
P
a
p
a
c
o
n
s
t
a
n
t
i
n
o
u
,
a
n
d
D
.
B
r
o
u
z
a
s
,
“
R
e
t
i
n
a
l
p
i
g
m
e
n
t
e
p
i
t
h
e
l
i
a
l
d
e
t
a
c
h
m
e
n
t
i
n
a
g
e
-
r
e
l
a
t
e
d
m
a
c
u
l
a
r
d
e
g
e
n
e
r
a
t
i
o
n
,
”
O
p
h
t
h
a
l
m
o
l
o
g
y
a
n
d
T
h
e
r
a
p
y
,
v
o
l
.
9
,
n
o
.
4
,
p
p
.
7
3
9
–
7
5
6
,
A
u
g
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
0
1
2
3
-
0
2
0
-
0
0
2
9
1
-
5.
[
1
2
]
A
.
G
a
r
c
í
a
-
F
l
o
r
i
a
n
o
,
Á
.
F
e
r
r
e
i
r
a
-
S
a
n
t
i
a
g
o
,
O
.
C
a
m
a
c
h
o
-
N
i
e
t
o
,
a
n
d
C
.
Y
á
ñ
e
z
-
M
á
r
q
u
e
z
,
“
A
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
t
o
m
e
d
i
c
a
l
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
:
D
e
t
e
c
t
i
n
g
a
g
e
-
r
e
l
a
t
e
d
m
a
c
u
l
a
r
d
e
g
e
n
e
r
a
t
i
o
n
i
n
f
u
n
d
u
s
i
m
a
g
e
s,
”
C
o
m
p
u
t
e
rs
&
El
e
c
t
ri
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
7
5
,
p
p
.
2
1
8
–
2
2
9
,
M
a
y
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
e
l
e
c
e
n
g
.
2
0
1
7
.
1
1
.
0
0
8
.
[
1
3
]
S
.
La
t
h
a
,
D
.
S
a
m
i
a
p
p
a
n
,
a
n
d
R
.
K
u
m
a
r
,
“
C
a
r
o
t
i
d
a
r
t
e
r
y
u
l
t
r
a
s
o
u
n
d
i
ma
g
e
a
n
a
l
y
s
i
s:
A
r
e
v
i
e
w
o
f
t
h
e
l
i
t
e
r
a
t
u
r
e
,
”
P
ro
c
e
e
d
i
n
g
s o
f
t
h
e
I
n
st
i
t
u
t
i
o
n
o
f
M
e
c
h
a
n
i
c
a
l
En
g
i
n
e
e
rs,
Pa
rt
H
:
J
o
u
r
n
a
l
o
f
E
n
g
i
n
e
e
r
i
n
g
i
n
M
e
d
i
c
i
n
e
,
v
o
l
.
2
3
4
,
n
o
.
5
,
p
p
.
4
1
7
–
4
4
3
,
Jan
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
7
7
/
0
9
5
4
4
1
1
9
1
9
9
0
0
7
2
0
.
[
1
4
]
J.
E.
W
.
K
o
h
e
t
a
l
.
,
“
D
i
a
g
n
o
s
i
s
o
f
r
e
t
i
n
a
l
h
e
a
l
t
h
i
n
d
i
g
i
t
a
l
f
u
n
d
u
s
i
m
a
g
e
s
u
si
n
g
c
o
n
t
i
n
u
o
u
s
w
a
v
e
l
e
t
t
r
a
n
sf
o
r
m
(
C
W
T)
a
n
d
e
n
t
r
o
p
i
e
s,”
C
o
m
p
u
t
e
rs
i
n
B
i
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
8
4
,
p
p
.
8
9
–
9
7
,
M
a
y
2
0
1
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
b
i
o
me
d
.
2
0
1
7
.
0
3
.
0
0
8
.
[
1
5
]
H
.
K
u
r
e
sa
n
,
D
.
S
a
m
i
a
p
p
a
n
,
a
n
d
S
.
M
a
s
u
n
d
a
,
“
F
u
si
o
n
o
f
W
P
T
a
n
d
M
F
C
C
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
i
n
P
a
r
k
i
n
s
o
n
’
s
d
i
sea
s
e
d
i
a
g
n
o
s
i
s,”
T
e
c
h
n
o
l
o
g
y
a
n
d
H
e
a
l
t
h
C
a
re
,
v
o
l
.
2
7
,
n
o
.
4
,
p
p
.
3
6
3
–
3
7
2
,
2
0
1
9
,
d
o
i
:
1
0
.
3
2
3
3
/
TH
C
-
1
8
1
3
0
6
.
[
1
6
]
R
.
M
o
n
i
k
a
,
S
.
D
h
a
n
a
l
a
k
sh
m
i
,
a
n
d
S
.
S
r
e
e
j
i
t
h
,
“
C
o
e
f
f
i
c
i
e
n
t
r
a
n
d
o
m
p
e
r
m
u
t
a
t
i
o
n
-
b
a
se
d
c
o
m
p
r
e
ss
e
d
se
n
si
n
g
f
o
r
med
i
c
a
l
i
ma
g
e
c
o
m
p
r
e
ss
i
o
n
,
”
i
n
A
d
v
a
n
c
e
s
i
n
E
l
e
c
t
ro
n
i
c
s,
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
C
o
m
p
u
t
i
n
g
,
S
p
r
i
n
g
e
r
S
i
n
g
a
p
o
r
e
,
2
0
1
7
,
p
p
.
5
2
9
–
536
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
981
-
10
-
4
7
6
5
-
7
_
5
6
.
[
1
7
]
D
.
S
a
m
i
a
p
p
a
n
a
n
d
V
.
C
h
a
k
r
a
p
a
n
i
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
u
l
t
r
a
so
u
n
d
c
a
r
o
t
i
d
a
r
t
e
r
y
i
ma
g
e
s
u
si
n
g
t
e
x
t
u
r
e
f
e
a
t
u
r
e
s,”
I
n
t
e
r
n
a
t
i
o
n
a
l
Re
v
i
e
w
o
n
C
o
m
p
u
t
e
rs
a
n
d
S
o
f
t
w
a
re
,
v
o
l
.
8
,
n
o
.
4
,
p
p
.
9
3
3
–
9
4
0
,
2
0
1
3
.
[
1
8
]
V
.
R
a
j
i
n
i
k
a
n
t
h
e
t
a
l
.
,
“
A
u
t
o
mat
e
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
r
e
t
i
n
a
l
i
ma
g
e
s
i
n
t
o
A
M
D
/
n
o
n
-
A
M
D
C
l
a
ss
—
a
s
t
u
d
y
u
si
n
g
m
u
l
t
i
-
t
h
r
e
s
h
o
l
d
a
n
d
G
a
ssi
a
n
-
f
i
l
t
e
r
e
n
h
a
n
c
e
d
i
ma
g
e
s,
”
Ev
o
l
u
t
i
o
n
a
r
y
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
4
,
n
o
.
2
,
p
p
.
1
1
6
3
–
1
1
7
1
,
M
a
r
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
2
0
6
5
-
021
-
0
0
5
8
1
-
2.
[
1
9
]
M
.
K
.
G
a
r
v
i
n
,
M
.
D
.
A
b
r
a
mo
f
f
,
R
.
K
a
r
d
o
n
,
S
.
R
.
R
u
sse
l
l
,
X
.
W
u
,
a
n
d
M
.
S
o
n
k
a
,
“
I
n
t
r
a
-
r
e
t
i
n
a
l
l
a
y
e
r
s
e
g
me
n
t
a
t
i
o
n
o
f
3
D
o
p
t
i
c
a
l
c
o
h
e
r
e
n
c
e
t
o
mo
g
r
a
p
h
y
u
s
i
n
g
c
o
a
r
s
e
g
r
a
i
n
e
d
d
i
f
f
u
si
o
n
m
a
p
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
Me
d
i
c
a
l
I
m
a
g
i
n
g
,
v
o
l
.
2
7
,
n
o
.
1
0
,
p
p
.
1
4
9
5
–
1
5
0
5
,
O
c
t
.
2
0
0
8
,
d
o
i
:
1
0
.
1
1
0
9
/
t
m
i
.
2
0
0
8
.
9
2
3
9
6
6
.
[
2
0
]
M
.
K
.
G
a
r
v
i
n
,
M
.
D
.
A
b
r
a
m
o
f
f
,
X
.
W
u
,
S
.
R
.
R
u
ss
e
l
l
,
T.
L.
B
u
r
n
s
,
a
n
d
M
.
S
o
n
k
a
,
“
A
u
t
o
ma
t
e
d
3
-
D
i
n
t
r
a
r
e
t
i
n
a
l
l
a
y
e
r
se
g
me
n
t
a
t
i
o
n
o
f
m
a
c
u
l
a
r
s
p
e
c
t
r
a
l
-
d
o
ma
i
n
o
p
t
i
c
a
l
c
o
h
e
r
e
n
c
e
t
o
mo
g
r
a
p
h
y
i
m
a
g
e
s,
”
I
E
E
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
Me
d
i
c
a
l
I
m
a
g
i
n
g
,
v
o
l
.
2
8
,
n
o
.
9
,
Evaluation Warning : The document was created with Spire.PDF for Python.