Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
C
om
put
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
8
,
No.
6
,
D
ece
m
ber
201
8
, pp.
5050
~
50
60
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v8
i
6
.
pp
5050
-
50
60
5050
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Optic Di
sc and
Macula
Locali
zation fro
m Retin
al Optic
al
Coher
ence Tom
ograph
y an
d Fun
dus Image
Rodiah
1
,
S
arif
uddin
M
ad
en
da
2
,
Diana
Tri
Su
se
tian
i
gt
i
as
3
, D
ew
i A
gush
inta R
ahayu
4
, E
ty
Sutan
t
y
5
1,3,4,5
Depa
rtment
of
Inform
at
i
cs
,
Gunada
rm
a
Uni
ver
sit
y
,
Indone
si
a
2
Doctor
al Progra
m
in
Inform
at
io
n
Technol
og
y
,
G
unada
rm
a
Univ
e
rsit
y
,
Indon
esia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
ul
22
, 2
01
8
Re
vised
Oct
3
,
201
8
Accepte
d
Oct
25
,
201
8
Thi
s
rese
ar
ch
u
sed
images
from
Optic
al
Coh
ere
nc
e
Tomogra
ph
y
(OCT)
exa
m
ina
t
ion
as
well
as
fundus i
m
age
s
to
localiz
e
the
optica
l
disc
and
m
ac
ul
ar
lay
er
of
retina
.
The
r
ese
ar
che
r
s
uti
lized
the
OCT
and
fund
us
image
to
int
erp
r
et
th
e
dis
ta
nc
e
be
twee
n
m
ac
u
la
r
ce
nt
er
and
optic
disc
i
n
the
imag
e.
The
dist
ance
wil
l
expr
ess
the
area
of
m
ac
u
la
that
ca
n
b
e
emplo
y
e
d
for
furth
er
rese
arc
h
.
Th
is
dista
nc
e
coul
d
r
e
cogni
z
e
the
thic
kness
of
m
ac
ula
par
ameters
dia
m
et
er
tha
t
wil
l
be
used
in
loca
li
zi
ng
p
roc
ess
of
opti
c
d
isc
and
m
ac
ula
.
The
par
amete
rs
are
t
he
ci
r
cle
rad
ius,
the
si
ze
of
wi
ndow’s
fil
te
r
,
t
he
consta
n
t
val
ue
and
the
s
iz
e
o
f
optic
d
isc
elem
ent
stru
cture
as
we
ll
as
the
siz
e
of
m
ac
ula
.
Th
e
res
ult
s
of
th
is
stud
y
are
exp
ec
t
ed
t
o
improve
th
e
a
cc
ura
c
y
o
f
m
ac
ula
d
etec
t
ion
that
expe
ri
ence the ede
m
a.
Ke
yw
or
d:
Fundus
Ma
cula
OCT
Op
ti
c
Disc
Re
ti
nal
Copyright
©
201
8
Instit
ute of
Ad
v
ance
d
Engi
ne
eri
ng
and
Sc
ie
n
ce
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Rod
ia
h,
Dep
a
rtm
ent
of
Inform
at
ic
s,
Guna
dar
m
a U
ni
ver
sit
y,
Ma
rgonda
Ra
y
a Street
100, P
ondok C
ina,
Depok,
In
donesi
a.
Em
a
il
: {ro
diah
, s
ari
f,
diants,
dew
ia
r,
ety
_s
}
@staff
.gu
nad
a
rm
a.ac.id
1.
INTROD
U
CTION
On
e
of
the
i
ndic
at
or
s
for
dia
betic
reti
no
pathy
in
adv
a
nce
sta
ge
is
ede
m
a
m
acula.
It
is
a
le
akag
e
of
fluid
s
in
reti
na.
Dia
betes
can
cause
sm
al
l
blo
od
vessel
i
n
t
his
area
to
rele
ase
flui
d
or
fat.
Edem
a
m
acula
m
a
y
aff
ect
the
patie
nt’s
visio
n
ce
nt
er
an
d
the
w
orst
-
case
ed
em
a
m
acula
m
ay
ca
us
e
blindness
[
1].
Ma
cula
is
a
part
of
reti
na
that
a
ll
ow
s
ey
e
to
se
e
so
ft
detai
ls
with
a
s
pheric
al
m
or
phol
og
y
that
locat
ed
i
n
the
bac
k
of
re
ti
na
at
3.5
m
m
tem
po
ral
distance
a
nd
0.5
m
m
s
m
a
ll
er
to
the
disc
us
[
2].
I
n
the
m
acula,
there
are
ves
sel
s
tha
t
play
i
m
po
rtant
r
ole
as
ph
ot
or
ece
pt
or
.
T
he
stu
dy
in
sh
ar
pen
i
ng
the
reti
nal
fundus
im
ages
was
pe
rf
or
m
ed
by
the
researc
hers
us
i
ng
a
dap
ti
ve
hi
stogram
and
m
orp
ho
l
og
ic
al
operati
on
of
to
phat
as
well
as
bo
t
hat.
Bl
ood
vessel
segm
entat
ion
i
s
n
ece
ssary
to
perform
in
local
iz
ing
oth
e
r
obj
ect
s
of
reti
na
l
fu
ndus
i
m
ages.
These
im
age
s
are
op
ti
c
disc
a
nd
m
acula
[3
]
.
I
n
a
ddit
ion
,
reti
nal
vasc
ular
e
xtracti
on
al
s
o
perfor
m
ed
us
i
ng
c
urvelet
trans
form
ation
an
d
m
or
ph
ologica
l
operati
on
that
com
bin
e
two
operati
ons
.
T
ho
se
ope
rati
on
s
are
t
ophat
da
n
bo
t
hat
[
4].
Th
e
local
iz
at
ion
of
t
he
opti
c
di
sc
an
d
m
acula
area
m
us
t
be
seper
at
e
d
f
rom
the
blo
od
ve
ssel.
Ther
e
f
or
e,
the
sh
ar
pness
of
hum
an
vison
is
highly
de
pend
ent
on
this
sect
ion
[5
]
.
T
he
m
acula
li
es
in
th
e
m
id
-
po
ste
rio
r
of
ret
ina.
On
e
of
the
init
ia
l
ph
ase
in
pe
rfor
m
ing
m
acula
detect
ion
th
at
exp
e
rience
d
edem
a
is
by
local
iz
ing
the
area
of
op
ti
c
disc
an
d
m
acula
area
of
reti
nal
im
age.
Opti
c
disc
local
iz
at
ion
in
this
s
tud
y
was
car
ri
ed
on
thr
ough
ta
gg
i
ng
the
co
nnect
ed
com
ponen
t
base
d
on
colo
r
intensit
y
that
hav
e
sim
il
ariti
es
with
ex
ud
at
es
by
check
i
ng
pix
el
(P
)
pe
r
pi
xel
(f
r
om
le
ft
to
righ
t
side
a
nd
from
top
to
bo
tt
om
directi
on)
to
identif
y
the
connecte
d
pixe
ls
area.
T
he
connecte
d
pixe
l
area
is
an
area
f
ro
m
the
bordere
d
pi
xe
l
that
hav
e
t
he
sam
e
intensit
y
or
int
ensity
value
in
a
set
of
V
(in
bin
ary
im
age
V
=
{1}
).
T
he
connecte
d
c
om
po
ne
nt
for
input
i
m
age
fo
ll
ow
s
the
connecti
vity
ru
le
s
(4
,
8
or
m
-
connecti
vity
)
[
6
]
.
This
stud
y
pro
po
se
s
a
new
te
chn
i
qu
e
t
o
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Op
ti
c
Disc
and Mac
ula L
oca
li
za
ti
on fro
m Re
ti
na
l O
ptical C
oh
e
rence
T
omog
r
aphy
and
F
undus
…
(
R
od
i
ah
)
5051
analy
ze
the
s
w
el
li
ng
of
the
m
acula.
T
his
stu
dy
co
ns
ist
s
of
t
wo
sta
ges.
The
first
sta
ge
is
ve
ssel
extracti
on
an
d
the sec
ond o
ne
is v
ess
el
r
ec
onst
ru
ct
io
n.
The
im
age
inp
ut
is
a
fu
nd
us
gr
ee
n
cha
nn
el
i
m
age
that
is
conver
te
d
into
gray
scal
e
i
m
age.
I
n
this
stud
y,
S
pee
ded
Up
Ro
bust
Feat
ur
es
(
SU
RF
)
al
go
rithm
was
us
ed
to
de
scri
be
the
proces
s
of
m
acula
detect
ion
on
reti
nal
i
m
age
by
discov
e
ri
ng
the
c
om
patib
il
ity
between
the
featur
e
s
po
ints
of
each
im
age.
T
his
study
al
so
us
e
d
Ra
ndom
Sam
ple
and
C
on
s
ens
us
(
RA
NSAC)
al
gorithm
to
m
a
tc
h
the
points
t
hat
co
rr
es
ponde
nt
a
nd
Gau
s
sia
n
f
un
ct
ion
al
gorithm
for
sm
oo
thi
ng
the
m
acular
im
age
sp
li
ci
ng.
The
RA
NSA
C
al
gorithm
that
w
a
s
i
m
ple
m
ented
i
n
this
stu
dy
use
d
to
sel
ect
ra
ndom
su
bs
et
f
ro
m
or
iginal
i
m
age
to
achie
ve
the
best
res
ults
in
local
iz
ing
the
s
wo
ll
en
m
acula
area
[
7
]
.
T
his
stud
y
determ
ined
the
pr
es
enc
e
of
m
acula
ed
e
m
a
by
detect
ing
t
he
pr
ese
nce
of
f
ovea.
F
ovea
li
es
in
the
m
idd
le
of
m
acula.
Fov
ea
and
op
ti
c
di
sc
detect
ion
w
ere
pe
rfor
m
ed
us
in
g
m
ot
ion
patte
r
n
recogn
it
io
n.
T
he
featu
re
sel
e
ct
ion
was
ca
rr
i
ed
on
usi
ng
Pri
ncipal
Com
ponen
t
An
al
ysi
s
(
PCA)
by
creati
ng
ne
w
featu
re
de
riv
ed
from
extracti
on
of
li
near
f
eat
ur
es
com
bina
ti
on
s.
T
her
e
f
ore,
they
can
be
us
ed
to
re
duce
t
he
s
el
ect
ed
feat
ur
e
dim
ension
w
hi
ch
is
e
dem
a
m
acula.
Ga
us
s
ia
n
de
rivati
ve
al
so
us
ed
in
th
is
stud
y
to cap
t
ur
e
the c
har
act
erist
ic
of edem
a
m
acula patt
ers
t
hat
will
b
e d
et
ect
e
d
[
8
]
.
The
reti
nal
f
undus
op
ti
c
dis
c
locat
iz
at
ion
was
pe
rfor
m
ed
in
se
ver
al
ste
ps
.
T
he
fir
st
ste
p
was
perform
ing
gr
e
en
cha
nel
im
age
fun
du
s
e
xtra
ct
ion
.
T
he
nex
t
ste
p
was
ca
rr
y
ou
t
fitur
op
ti
c
nerve
head
(ONH)
.
ONH
was
perform
ed
based
on
the
c
on
t
ou
r
f
ro
m
op
ti
c
disc
cu
p
a
rea.
In
pe
rfor
m
ing
the
e
xtracti
o
n,
the
researc
hers
c
onve
rtin
g
the
im
age
to
gr
ay
sc
al
e
and
e
xecu
ti
ng
th
resholdi
ng
with
dilat
ion
and
er
otio
n
operati
on.
The
ne
xt
ste
p
is
perform
ing
edg
e
detect
io
n
us
in
g
dilat
ion
operati
on
t
o
c
reate
m
at
rix
featur
e.
The
res
ult
of
m
at
rix
featu
re
will
b
e
us
ed
as
an
i
nput fo
r
tr
ai
nin
g an
d
te
sti
ng pro
c
ess.
Th
e trai
ning a
nd t
est
ing
process
in this
stud
y
us
i
ng
S
VM
m
et
ho
d
[
9]
.
The
researc
h
was
perform
e
d
by
est
ablishi
ng
a
n
autom
at
i
c
diagnostic
syst
e
m
to
detect
m
acule
us
in
g
flu
oresce
in
i
m
age
.
In
th
e
pr
el
im
inary
sta
ges,
n
oise
re
du
ct
io
n
wa
s
done
in
pr
e
proc
essing
ph
a
se
to
el
im
i
nate
the
obj
ec
t
detai
ls
besid
e
the
m
acula
and
t
o
m
ai
nta
in
it
s
li
gh
tnin
g
distri
bu
ti
on
wit
h
Gau
s
sia
n
filt
er
.
The
re
we
re
sever
al
proces
ses
in
this
stu
dy,
f
or
in
sta
nc
e
blood
vessel
segm
entat
ion
with
Gau
s
si
an fil
te
r, o
ptic disc
dete
ct
ion
, a
nd
red le
sion detec
ti
on
to
se
gm
ent
m
i
cro
ane
ur
ysm
f
eat
ur
e.
The
segm
ented
featu
re
will
descr
i
be
the
sh
ape
,
colo
r
an
d
co
ntra
st
of
t
he
detect
ed
ca
nd
i
dates
[
10
].
The
opti
c
disc
detect
ion
in
t
his
stud
y
was
do
ne
by
detect
in
g
th
e
m
axi
m
um
area
with
the
highest
pix
el
value.
The
n
the
reti
na
l
fun
du
s
im
age
al
ong
with
the
opti
c
disc
that
al
read
y
bei
ng
detect
ed
will
pe
rfor
m
local
im
age
featur
e
s
pr
oces
s.
Pixel
weig
ht
ing
was
ca
rr
ie
d
out
to
gai
n
ne
ighborin
g
in
fl
uen
ce
in
perform
ing
segm
enta
ti
on
by
cl
us
te
rin
g
e
ach
pi
xel
insid
e
the
cl
us
te
r
c
enter.
T
he
m
a
cula
cl
assifi
cat
ion
ph
ase
pro
pose
d
in
this
st
ud
y
is
base
d
on
s
om
e
par
am
et
ers
suc
h
as
the
siz
e
of
m
acula
window’
s
area
a
nd
sp
at
ia
l
relat
ion
s
hi
p
bet
wee
n
pi
xels
in the ret
inal
fundus
[
11
].
The
m
edical
e
xam
inati
on
of
the
edem
at
ou
s
m
acules
is
done
by
direct
obs
erv
at
io
n
by
the
ph
ysi
ci
a
n
thr
ough
O
ptica
l
Coh
ere
nce
T
om
og
ra
ph
y
(
O
CT)
exam
inatio
n.
This
e
xam
i
nation
will
pro
vid
e
cr
os
s
sect
ion
al
of
reti
nal
iss
ue
i
m
ages.
I
n
thi
s
exam
inati
on
,
the
thic
knes
s
of
the
reti
na
a
nd
m
acula
will
be
m
easur
e
d,
wh
ic
h
ind
ic
at
e
the
oc
currence
of
sw
el
li
ng
an
d
flui
d
cl
ots
in
the
a
rea.
Dete
rm
ina
ti
on
of
opti
c
di
sc
local
iz
at
ion
in
this
stud
y
car
ried
on
us
i
ng
fun
dus
i
m
age.
Disc
local
iz
at
ion
an
d
m
acula
are
pr
incipal
fact
or
s
tha
t
determ
ined
the
distance
i
nter
pret
at
ion
s
betw
een
the
cente
r
s
of
m
acula
w
it
h
opti
c
disc
i
n
pix
el
repres
entat
ion
t
o
kn
ow
the
diam
e
te
r
of
m
acular
thicke
ni
ng.
This
stu
dy
us
e
d
so
m
e
par
a
m
et
ers.
Fo
r
in
sta
nce,
the
ci
rc
le
rad
ius,
t
he
siz
e
of
window
filt
er,
and
c
on
sta
nt
va
lue
as
well
as
the
siz
e
of
op
t
ic
disc
structuri
ng
el
em
ent
an
d
m
acula.
The
resu
lt
of this st
ud
y i
s
exp
ect
e
d
t
o
im
pro
ve
the
accu
racy o
f
m
acular
detect
ion
t
hat experie
nced e
dem
a.
2.
RESEA
R
CH MET
HO
D
2.1.
Data Se
ts
This
stu
dy
us
ed
im
aging
im
ages
with
O
ptica
l
Coh
e
rence
Tom
og
ra
ph
y
(O
CT)
an
d
Fluo
re
scei
n
An
giograp
hy
f
undus im
age as
can
b
e
seen
in
Fi
gure
1.
Fig
ure
1. OCT
and F
undus
Im
age
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5050
-
5060
5052
In
t
he
f
undus
reti
nal
i
m
age
in
Fig
ure
1,
t
he
i
m
age
sh
ow
s
qu
it
e
lo
w
c
ontrast
of
reti
na
’s
m
acula.
Ma
cula
ob
j
ect
s
are
s
om
eti
m
e
hi
nd
e
red
by
ano
the
r
obj
ec
t
that
ind
ic
at
e
reti
nal
abn
or
m
al
i
ti
es
su
ch
as
the
exudate,
dot
a
nd
blood
hem
orrh
a
ge,
m
ic
ro
aneurysm
as
well
as
In
tra
-
Re
ti
nal
Mi
cro
vasc
ular
A
bnorm
aliti
es
(I
RM
A
’s). T
he
se ab
norm
al
itie
s can
cause
d
i
f
ficult
ie
s and i
n l
ocali
zat
ion
of m
acular ar
ea.
2.2.
Estima
te Opti
c D
isc
P
os
iti
on
In
Flu
or
escei
n
Angiog
raphy
(
FA
)
f
undus
im
age,
t
he
init
ia
l
ph
a
ses
of
t
he
proces
s
we
re
de
te
ct
ing
th
e
center
e
stim
at
i
on
an
d
the d
ia
m
et
er
of
opti
c d
isc
. Th
e
est
im
at
ion
o
f
m
acula
area
was
ba
s
ed
on
the
dista
nce
a
nd
it
s
po
sit
io
n
with
op
ti
c
disc.
T
his
distance
is
relat
ively
co
nst
ant
in
each
r
et
ina.
T
he
siz
e
of
the
disc
on
eac
h
reti
na
im
age
of
the
pati
e
nt
is
va
ries,
but
ge
ner
al
ly
the
siz
e
of
op
ti
c
disc
has
diam
et
er
betwee
n
1.8
±
2mm
inside
the
im
a
ge
fun
dus
with
sta
nd
ar
d
siz
e
of
76
8
x
57
6
pi
xel.
Ther
e
a
re
so
m
e
factor
s
that
m
ake
op
ti
c
disc
edg
e
detect
io
n
accurate.
Som
e
par
ts
of
t
he
op
ti
c
disc
w
il
l
be
ob
sc
ur
e
d
by
m
any
reti
nal
blood
ves
sel
s
.
In
gen
e
ral,
the
process
in
l
ocali
zi
ng
the
e
xiste
nce
of
the
op
ti
c
disc
in
this
st
ud
y
was
pe
rform
ed
by
the
f
ollow
in
g
al
gorithm
:
1.
Gr
ee
n
c
ha
nn
el
FA
f
undus
im
age
e
xtracti
on.
The
gr
ee
n
c
ha
nn
el
was
sel
e
ct
ed
as
the
pri
m
ary
colo
r
si
nc
e
gr
ee
n
c
olor
ha
s
the
high
i
ntensity
com
par
e
to
the
oth
e
rs
t
wo
col
or
s;
re
d
and
bl
ue
in
a
n
i
m
age
with
R
GB)
colo
r
as ca
n be
seen
in Fi
gure
2
.
Bel
ow is th
e m
a
them
a
ti
cal
f
or
m
ula to f
i
nd the
gree
n
c
ha
nn
el
sel
ect
ion
:
)
(
B
G
R
G
g
(1)
Fig
ure
2. OCT
and Gree
n
Cha
nn
el
Fun
dus Fl
uorescein
An
gio
gra
ph
y
(FA)
Im
age
2.3.
Mo
r
f
ology
Fo
rm
ula
(1)
is
perform
ed
to
find
a
gr
e
en
c
ha
nn
el
in
an
im
age,
w
hile
R,
G
an
d
B
are
sta
nd
f
or
Re
d,
Gr
ee
n
a
nd
Bl
ue
seq
uen
ti
al
ly
.
In
gr
e
en
c
ha
nnel
,
the
detai
l
of
i
m
age
can
be
seen
cl
early
an
d
com
pr
e
hensi
vely
.
Wh
e
reas
t
he
use
d
of
re
d
c
ha
nn
el
only
goin
g
to
s
how
t
he
i
m
age
ed
ge
w
hile
in
blu
e
ch
ann
el
m
erely
pro
vid
e
j
ust
p
art of the
i
m
age’
s d
et
ai
l and
c
onta
in a lot o
f n
oise. Ba
sed
on im
age i
n
Fig
ur
e
2,
t
he
con
t
rast o
f o
ptic disc
can
be
obser
ve
d
bette
r
i
n
green
c
ha
nn
el
c
om
par
ed
t
o
re
d
a
nd
blu
e
c
ha
nn
el
in
R
GB
i
m
age.
The
refor
e
,
thi
s
stud
y
on
ly
us
e
d
gree
n
cha
nn
el
i
m
age
to
co
m
pu
te
the
op
ti
m
al
threshold.
Figure
3
sho
ws
the
gr
ee
n
c
hannel
fun
du
s
Fl
uore
scei
n
Angiog
r
aph
y
(FA
)
im
age
and
it
s
histogram
.
It
can
be
seen
that
the
pix
el
that
corres
pondin
g wit
h op
ti
c
disc
hav
e
h
i
gh
e
r
i
ntensity
in hist
ogram
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Op
ti
c
Disc
and Mac
ula L
oca
li
za
ti
on fro
m Re
ti
na
l O
ptical C
oh
e
rence
T
omog
r
aphy
and
F
undus
…
(
R
od
i
ah
)
5053
Fig
ur
e
3
.
OCT
an
d Gr
ee
n
C
ha
nn
el
Fun
dus
Fluorescei
n A
ngio
gr
a
phy (F
A
)
Im
age
2.
The
local
iz
at
ion
of
opti
c
di
sc
posit
ion.
O
ptic
disc
local
iz
at
ion
can
be
perform
ed
by
consi
de
rin
g
th
e
geo
m
et
ry
sh
ap
e
of
op
ti
c
disc
on
se
ve
ral
co
nn
ect
e
d
com
po
ne
nts
.
T
he
diff
ic
ulti
es
in
est
i
m
ating
the
opti
c
disc
posit
ion
is
the
presence
of
ot
her
obj
ect
s
ou
tsi
de
the
op
t
ic
disc
i
m
age
that
has
c
onnec
te
d
com
po
ne
nt
as w
el
l as ca
n be see
n
i
n
Fi
gure
4.
Fig
ure
4
.
Co
nnect
ed
Com
pone
nt D
et
ect
io
n o
n
Op
ti
c
Disc
3.
Each
com
pone
nt
that
conn
ect
ed
to
the
im
age
will
be
la
belle
d
with
the
nu
m
b
er
of
pix
el
s
in
the
com
po
ne
nt.
S
patia
l
coord
i
na
te
of
each
c
on
nected
c
om
po
nen
t
is
cal
cula
te
d.
Com
po
ne
nt
is
assum
ed
to
hav
e
m
axi
m
um
nu
m
ber
of
pi
xels
with
m
axi
m
u
m
dia
m
et
er
of
op
ti
c
disc
wh
ic
h
is
2
m
m.
Co
ns
e
quently
,
if
on
e
com
ponent
of
t
he
im
age
sp
at
ia
l
co
ordin
at
e
is
within
50
distance
pix
e
ls
from
sp
at
ia
l
coor
din
at
e
of
t
he
m
ean
la
rg
est
com
po
ne
nt,
the
obj
ect
will
be
com
bin
ed
an
d
the
new
m
ean
of
im
age
sp
at
ia
l
coo
r
din
at
e
will
be
c
om
pu
te
d
a
s can
b
e
seen
in
Fi
gure
5.
Fig
ure
5
.
P
os
it
ion Est
im
a
ti
on
of
O
ptic Disc
4.
Perfo
rm
ing
conn
ect
e
d
c
om
ponen
t t
a
gg
i
ng wi
th obj
ect
g
e
om
et
ry app
r
oac
h usin
g
s
om
e
par
am
et
ers
belo
w
:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5050
-
5060
5054
1)
The val
ue
area
is com
pu
te
d ba
sed o
n
the
m
axi
m
u
m
an
d
m
ean in
te
ns
it
y i
n
t
he op
ti
c
disc a
r
ea.
2)
Re
stric
ti
ng
the
area
that
are
detect
ed
as
an
op
ti
c
disc.
T
his
area
is
the
nu
m
ber
of
im
ag
e
di
m
ension
with a
ng
le
a
nd
width pa
ram
eter
s in
r
et
inal i
m
age th
at
is
wri
tt
en
in
a
n
ar
ra
y form
u
sing t
he
for
m
ula
:
Boun
dingBo
x=
[corn
e
r widt
h]
(2)
3)
Re
ti
nal o
ptic disc po
sit
ion
est
i
m
ation
b
ase
d on
the
rati
o
w
he
re th
e sh
a
pe
of the opti
c d
isc
is circl
e. I
n
a
ci
rcle
obj
ect
the
pro
per
ti
es
t
hat
can
be
disc
ov
e
re
d
a
re
[
12
]
:
Ra
diu
s
of
th
e
ci
rcle
r
,
ci
rcle
diam
e
te
r
d
,
wi
de
of
ci
r
cl
e
L
,
ci
rc
um
fer
en
ce
of
ci
rcle
a
nd
P
par
am
et
er.
The
f
orm
ula
to
local
iz
e
op
ti
c
dis
c
po
sit
io
n base
d on the
rati
o o
f
rou
nd
4)
The
e
xtensi
ve
an
d
pa
ram
et
e
r
val
ue
as
t
he
pro
per
ty
of
t
his
ci
rcle
ca
n
be
perform
ed
in
opti
c
disc
reg
i
on
s
which
will
b
e loc
al
iz
ed
as
the
basic
f
or
m
o
f
t
he ro
unda
bout
siz
e usi
ng this
form
ula:
L
P
R
a
t
i
o
4
2
(3)
5)
If
t
he
resu
lt
va
lue
of
detect
ed
obj
ect
rati
o
yi
el
ds
1
a
nd
has
the
inten
sit
y
equ
al
or
al
m
os
t
the
sam
e
as
op
ti
c d
isc
inte
nsi
ty
, th
is detec
t
ed
obj
e
ct
w
il
l
be
m
ark
ed
(loc
al
iz
at
ion
).
T
he
obj
ect
s that h
a
ve
dif
fer
e
nt
char
act
e
risti
c
with
the
opti
c
disc
will
no
t
be
m
ark
ed.
In
ge
ner
al
,
the
op
t
ic
disc
il
lustrati
on
can
be
seen i
n
Fi
gure
6.
Fig
ure
6
.
O
ptic Disc P
os
it
io
n Local
iz
at
ion
Illustrat
io
n
The
rati
o
valu
e
of
1
is
base
d
on
t
he
rati
o
le
ng
t
h
of
m
ajo
r
axis
an
d
m
ino
r
axis
of
the
ob
je
ct
that
was
detect
ed
i
n
f
un
du
s
im
age as
can
be
see
n
i
n
F
igure
7.
Fig
ure
7
.
Ma
Detect
ed
Op
ti
c
D
isc
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Op
ti
c
Disc
and Mac
ula L
oca
li
za
ti
on fro
m Re
ti
na
l O
ptical C
oh
e
rence
T
omog
r
aphy
and
F
undus
…
(
R
od
i
ah
)
5055
2.4.
Localiz
at
ion
OCT
M
acula
Lay
ers
Locali
zat
ion
of
m
acular
i
m
a
ge
la
ye
r
a
rea
was
perform
ed
on
pr
e
process
ing
phase.
Thi
s
local
iz
at
io
n
was
ca
rr
ie
d
on
in
cal
culat
in
g
the
m
acular
ar
ea
outsi
de
t
he
norm
al
siz
e
of
reti
na
m
acular
i
m
age.
T
he
m
acula
area
has
the
thi
ckn
e
ss
of 4
00
μm
,
di
m
inishing
on f
ov
ea with
the
siz
e o
f
15
0
μm
,
and
bec
om
ing
thin
ner
o
n
ora
serr
at
a
with
th
e
thick
ness
of
80
μm
[13
]
.
The
m
acula
ha
s
ve
ry
low
c
on
t
rast
inside
the
reti
nal
im
a
ge
a
nd
so
m
et
i
m
e
the
area
of
m
acul
a
is
blo
cke
d
by
the
exudates
or
the
blee
di
ng
in
t
he
reti
na
’s
re
gion
[
1
4
]
.
The
al
gorithm
s o
f p
repro
ce
ssin
g p
hase
on
OCT
m
acular im
age
s ar
ea ca
n be
e
xp
la
ine
d
bel
ow
:
1.
Re
ad
the
O
CT
Im
age
2.
Perfo
rm
ing
the
blurrin
g
pro
cess
to
re
duce
the
noise
in
OCT
im
age.
The
inte
ns
it
y
value
of
the
im
a
ge
changes
a
lot
as
in
Fi
gure
8.
The
noise
in
O
CT
im
ages
ha
ve
the
f
or
m
of
dots
or
pixe
ls
with
dif
fer
e
nt
intensit
y.
This
stud
y
us
ed
the
first
de
rivati
ve
of
Ga
us
sia
n
filt
er
to
re
duce
the
noise
on
OCT
im
age
with
the form
ula
bel
ow
:
2
2
2
2
2
2
1
)
,
(
y
x
e
y
x
G
(4)
Wh
e
re
de
no
te
the
der
iv
at
ive
sta
nd
a
rd
f
r
om
pix
el
distrib
ution
of
OCT
im
age
OCT
)
,
(
y
x
I
.
Assum
ed
a
distrib
ution h
a
ve
a
n
a
ver
a
ge value e
qual
to
0 (
=1)
.
Fig
ure
8.
N
ois
e Red
uction Pr
ocess
a
nd
OCT
Mac
ula I
m
age
(a)
O
rigin
al
Im
age
(b)
Im
age f
ro
m
Bl
ur
rin
g P
ro
ces
s
3.
Perfo
rm
ing
ed
ge
detect
ion
on
im
age
gr
a
die
nt
at
ho
rizo
ntal
directi
on
f
r
om
the
filt
ering
i
m
age
us
i
ng
the
al
gorithm
s b
el
ow
:
a)
Re
ad
the
im
ag
e res
ulted
from
b
lu
rr
i
ng phase
.
b)
Re
ad
siz
e o
f
th
e i
m
age w
it
h
fu
nctio
n
[
N,
M]
=size
(
G)
.
[N
M]
is a p
ixel l
i
ne
an
d
col
um
n
of
an
im
age
ob
ta
ine
d
f
ro
m
blurrin
g
phase
G
(
x
,
y
)
.
The
fi
rst
der
ivati
ve
of
im
age
fu
nct
ion
G(
x
,
y
)
towa
rd
s
x
a
nd
y
axis th
at
a
re c
om
pu
te
d
us
i
ng for
m
ula [
15
]
:
)
,
(
y
x
x
G
an
d
)
,
(
y
x
y
G
(5)
Wh
e
reas,
the
ve
ct
or
gr
a
die
nt
and the am
plit
ud
e
can
b
e
calc
ulate
d wit
h [9
]
:
)
,
(
)
,
(
)
,
(
y
x
y
G
y
x
x
G
y
x
G
(6)
2
2
)
,
(
)
,
(
)
,
(
|
|
y
x
y
G
y
x
x
G
y
x
G
(7)
4.
Re
fer
ri
ng
t
o
e
qu
at
io
n
7,
the
first
pa
rtia
l
deri
vative
of
im
a
ges
f
r
om
blu
rr
i
ng
phase
w
ere
com
pu
te
d
ba
s
ed
on the t
wo
-
cl
ose
d neig
hbori
ng
pix
el
. T
he
refor
e
, th
e
gradie
nt can be
obta
i
ned
:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5050
-
5060
5056
)
,
(
)
,
1
(
)
,
(
)
,
(
y
x
G
y
x
G
y
x
x
G
y
x
G
x
(8)
x
y
x
G
)
,
(
is
an
i
m
age
gradie
nt
on
x
horiz
on
ta
l
direc
ti
on
that
pro
duced
e
dge
of
obj
ect
in
the
s
hap
e
of
horizo
ntal l
ine,
as in
Fig
ur
e
9.
Fig
ure
9
.
Im
age Grad
ie
nt on Hori
zo
ntal Dir
ect
ion
5.
Labeled
t
he
c
onnected
c
om
pone
nt
that
li
nk
ho
rizo
ntal
im
age
gr
a
dient
with
ad
j
ace
nt
m
at
rix.
x
pixe
l
i
s
sai
d
to
be
a
dja
cent
with
y
pi
xe
l
if
there
is
a
connecte
d
c
on
necti
on
betwee
n
them
.
Two
i
m
ages
subset
S
1
and
S
2
are a
dja
cent if s
om
e p
ixels
on
S
1
a
dja
cent with
s
om
e
S
1
pix
el
s a
s in
Figure
10.
Fig
ure
10
.
Sub
set
S
1
a
nd
S
2
Path
f
ro
m
pix
e
l
x
with
co
ordin
a
te
)
,
(
q
p
for
y
pi
xel
with
co
ordi
nate
)
,
(
s
r
is
a
pix
el
order
that
dif
fer
from
coor
din
at
e
)
,
)
,
.
.
.
(
,
(
),
,
(
1
1
0
n
n
o
q
p
q
p
q
p
wh
e
re:
n
= Path
)
,
(
)
,
(
),
,
(
)
,
(
0
0
s
r
q
p
q
p
q
p
n
n
(9)
)
,
(
1
1
y
p
is
adjace
nt
where :
1
1
,
i
i
q
p
(10)
6.
Com
pu
te
the
di
sta
nce
betwee
n
)
,
(
q
p
x
and
)
,
(
s
r
y
us
ing
ci
ty
blo
ck
distance
with
4
co
nn
ect
ivit
y
a
s
can
be
see
n
i
n
Figure
11
us
in
g
the
for
m
ula :
|
|
|
|
)
,
(
4
s
q
r
p
y
x
D
(11)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Op
ti
c
Disc
and Mac
ula L
oca
li
za
ti
on fro
m Re
ti
na
l O
ptical C
oh
e
rence
T
omog
r
aphy
and
F
undus
…
(
R
od
i
ah
)
5057
Fig
ure
11
.
Ci
ty
Bl
ock
Dista
nc
e w
it
h 4 C
onne
ct
ivit
y
3.
RESU
LT
S
A
ND AN
ALYSIS
3.1.
Experim
en
ta
l
Ima
ge
The
local
iz
at
ion
of
the
opti
cal
disc
posit
ion
was
perf
orm
ed
in
ge
ome
tric
op
ti
c
dis
c
form
by
consi
der
i
ng
se
ver
al
c
onnecte
d
com
ponen
ts
.
The
diff
ic
ulty
in
pe
rfor
m
ing
the
opti
c
disc
posit
ion
est
i
m
at
io
n
was
the
occ
urr
ence
of
a
nothe
r
ob
j
ect
outsi
de
the
op
ti
cal
di
sc
that
ha
s
a
c
onnected
com
po
ne
nt
as
well
.
Figur
e
12 sho
ws
t
he p
rop
os
ed
alg
or
it
hm
in
this
stu
dy
su
ccee
d
to
lo
cal
iz
e the pres
ence
of opti
c di
sc on fu
ndus i
m
age.
Figure
12.
E
sti
m
at
ion
I
m
age O
D
P
os
it
ion
In
pre
vious
stu
dy,
the
process
of
reti
nal
fun
du
s
op
ti
c
disc
i
m
ages
local
izati
on
was
pe
rfor
m
ed
us
i
ng
com
bin
at
ion
of
m
or
phol
og
ic
al
op
e
rati
on
a
nd
op
ti
c
disc
ce
nter
po
i
nt
de
te
rm
inati
on
.
Thi
s
m
e
thod
wa
s
carried
ou
t by
detect
in
g
the g
eom
et
ric
sh
a
pe
in
the p
aram
e
tric
fo
r
m
,
wh
ere
the
s
hap
e of
the o
pt
ic
disc
is
ci
rcu
l
ar.
T
he
exp
l
oitat
ion
of
Hough
T
ran
s
f
or
m
at
ion
[16]
was
car
ried
ou
t
to
search
the
m
os
t
su
it
able
geo
m
et
ric
fo
r
m
that
corres
pond
to
t
he
set
of
point
s
in
the
im
age.
The
t
ran
s
f
or
m
at
ion
is
done
t
o
fi
nd
the
li
n
e
s
an
d
the
cu
rv
e
s
that
passing
by
m
a
ny
points
i
ns
id
e
the
im
age.
These
li
nes
an
d
curves
a
re
the
on
e
that
near
a
nd
m
os
t
cor
re
s
pond
with
th
e
data
in
the
im
age
m
at
rix.
The
sel
e
ct
ion
of
the
ci
r
cular
ra
diu
s
gr
eat
ly
aff
ect
the
accu
racy
of
t
he
area
local
iz
at
ion
of
op
ti
c
disc
.
If
t
he
ci
rc
ular
pa
r
a
m
et
er
value
is
too
sm
al
l
or
too
big
,
it
will
aff
ect
the
distance
of
m
acula’s
cente
r
f
ro
m
the
ce
nt
ral
of
opti
c
di
sc,
res
ulti
ng
m
acula
RO
I
will
be
unsu
it
a
ble.
The
stu
dy
in
this
pap
e
r
us
ed
r
oundness
rati
o.
T
his
r
oundne
ss
rati
o
is
the
rati
o
of
the
ra
diu
s
at
the
sm
alles
t
ang
le
to
t
he
r
adiu
s
aver
a
ge
of
f
undus
im
age
as
obj
ect
.
The
r
at
io
of
t
his
st
ud
y
was
pro
pose
d
by
the
r
esearche
rs
t
o
perfor
m
com
par
ison
be
tween
the
s
hor
thest
axis
with
the
lo
ngest
a
xis
of
t
he
op
ti
c
di
sc.
Ta
ble
1
s
ho
ws
e
xam
ples
of
F
A
fun
du
s
I
m
age
OD esti
m
a
ti
on
po
sit
io
n.
As
ca
n
be
see
n
on
Ta
ble
1,
the
opti
c
disc
with
the
r
oundne
ss
obj
ect
r
at
io
app
r
oac
h
su
cces
fu
ll
y
local
iz
ing
it
s
posit
ion
c
orrect
ly
.
This
rati
o
is
the
ave
rag
e
geo
m
et
ric
diam
et
er
of
opti
c
disc
with
it
s
lon
gest
diam
e
te
r,
wh
e
r
e
the
obj
ect
with
a
ro
un
de
d
ge
om
et
ric
sh
ape
connecte
d
co
m
po
nen
t
in
f
undus
im
age
is
an
opti
c
disc. T
he
refor
e
, those a
rea a
re
locali
zed as
opti
c d
isc
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
87
08
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5050
-
5060
5058
Table
1.
O
ptic Disc L
ocali
zat
ion R
esult
I
m
ag
e
FA Fun
d
u
s I
m
ag
e
Op
tic Disc
Localiz
atio
n
Resu
lt
FA1
FA2
FA3
FA4
FA5
3.2.
OCT
M
acula
Ima
ge
L
oca
li
z
at
i
on
The
m
acula’s
la
ye
rs
play
i
m
portant
r
ole
in
hu
m
an’
s
vision.
T
hese
m
ul
ti
ple
la
ye
rs
are
locat
ed
i
n
three
dif
fer
e
nt
areas
(Fov
ea
, Avas
cular
Z
one an
d
F
oveola
r
)
as ca
n be
see
n
in
Fig
ure
13.
Figure
13.
(a)
Fundus
Sca
nn
e
d Im
age F
A, (
b).
Ar
ea
and
La
ye
rs
of
OCT M
acula,
(c)
.
Mac
ula
Area Po
sit
ion
Edem
a
m
acula
is
ass
ociat
ed
with
visio
n
blindness
in
non
prolife
rati
ve
diabeti
c
reti
nopathy.
Th
e
sy
m
pto
m
s
are
char
act
erize
d
by
(i)
the
reti
nal
thickeni
ng
at
or
within
500µm
distance
fr
om
the
center
of
m
acula,
(ii)
th
e
occurre
nce
of
ex
ud
at
es
or
within
500µm
distance
f
ro
m
t
he
center
of
m
acula
fo
ll
owe
d
by
the
thickeni
ng
of
t
he
reti
nal
surr
ound
the
m
acu
la
’s
center
an
d
(iii
)
reti
nal
thickenin
g
that
great
er
than
on
e
disc
diam
e
te
r
(D
D
)
distance
from
the
center
of
m
acula.
The
c
ha
ng
e
s
inside
the
m
acula’s
la
yer
area
will
influ
enc
e
the
per
ip
he
ral’
s
visi
on.
T
her
e
fore,
it
is
ve
ry
i
m
po
rtant
t
o
l
oc
al
iz
e
the
area
of
m
acula
in
de
t
ect
ing
m
acula
that
has
e
dem
a
[1
7
]
.
Fig
ure
14
s
hows
t
he
e
x
am
ple
of
OCT
m
a
cula
la
ye
rs
l
oc
al
iz
at
ion
’s
res
ul
ts
on
I
nner
Lim
iting
Mem
br
ane
(IL
M) area
and R
PE O
C
T m
acula im
age.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
& C
om
p
Eng
IS
S
N: 20
88
-
8708
Op
ti
c
Disc
and Mac
ula L
oca
li
za
ti
on fro
m Re
ti
na
l O
ptical C
oh
e
rence
T
omog
r
aphy
and
F
undus
…
(
R
od
i
ah
)
5059
Co
de
No
te
ILM
Inn
er
Li
m
itin
g
M
e
m
b
rane
NFL/
GCL
Nerve Fib
er
L
a
y
e
r
IPL
/
INL
Inn
er
Plex
if
o
r
m
La
y
er
INL
/
OPL
Inn
er
Nu
clea
r
L
ay
er
OPL/
ONL
Ou
ter
Plex
if
o
r
m
L
ay
er
OPL/
ONL
Ou
ter
Nu
clea
r
Layer
IS/
OS
Inn
er
Seg
m
en
t
RPE
Retin
al Pig
m
en
ted
Epith
elliu
m
Figure
14.
Ma
cula A
rea
L
ocal
iz
at
ion
a
nd
OC
T Mac
ula Lay
e
r
[
17]
Ma
ny
ap
proac
hes
ha
d
be
en
done
by
s
om
e
resea
rch
e
rs
i
n
OCT
m
acula
i
m
ages
local
iz
at
ion
.
This
local
iz
at
ion
w
as
car
ried
out
t
hro
ugh
sel
ect
i
ng
the
pi
xel
th
at
will
be
us
e
d
as
see
d
point
and
by
searc
hi
ng
the
si
m
il
ar
neighb
or
i
ng
pi
xels.
T
he
pix
el
s
a
rea
then
e
xpan
de
d
by
a
dd
i
ng
sim
il
ar
pix
el
t
o
th
e
neig
hbori
ng
pix
el
s
.
Unfortu
natel
ly
,
researc
hers
f
aci
ng
dif
ficult
ie
s
of
perfor
m
ing
segm
entat
ion
on
OC
T
m
aku
la
area.
Th
e
diff
ic
ulti
es
are
on
c
olor
sim
ilariti
es
and
on
te
xtu
re
of
eac
h
m
acula
la
ye
r.
Ma
cula
la
ye
r
s
co
ns
ist
of
8
l
ay
ers
OCT. In
this st
ud
y, t
he
resea
r
cher
s
pro
posed
the u
ti
li
zi
ng
of ad
j
ace
ncy m
a
trix to prese
nt the p
i
xels of m
acular
la
ye
r.
Th
ese la
ye
rs
are close
d t
o
the o
t
her
m
acula’s lay
er. The a
dj
ace
ncy
was
pe
rfo
rm
ed
by p
re
sentin
g pixel 1
wh
ic
h
in
dicat
ing i
f n
od
e
i t
o node
j
has
an e
dg
e
and
0 othe
rw
ise
(whe
n
t
he
nod
e
h
a
s
no
edg
e
).
4.
CONCL
US
I
O
N
The
pro
pose
d
al
gorithm
su
ccee
ds
to
local
iz
e
the
existe
nce
of
opti
c
disc.
This
opti
c
disc
is
based
on
the
co
nnect
ed
com
po
ne
nt
tha
t
disti
nguish
e
d
opti
c
disc
with
the
ot
her
ob
je
ct
of
t
he
im
age
fun
dus
s
uch
as
the
exudates.
T
he
resu
lt
of
m
acula’s
area
local
iz
at
ion
on
OC
T
i
m
age
can
co
m
pu
te
the
m
ac
ula
area
outsi
de
the
norm
al
siz
e
of
m
acula
reti
na
i
m
age.
The
a
ppr
oach
with
geo
m
et
ric
op
ti
c
disc
rati
o
s
uc
ceeds
to
local
iz
ed
the
op
ti
c
disc.
T
he
ex
plo
it
at
ion
of
ad
j
acen
dy
m
atr
ix
a
ble
to
disti
nguis
h
eve
ry
la
ye
r
of
OCT
m
a
cula
im
age
desp
it
e
their sim
il
ari
ti
e
s and c
olor
neighb
or
i
ng.
Fo
r
t
he
f
ur
the
r
resea
rch,
the
thicknes
s
of
7
la
ye
rs
of
m
acula
OCT
im
age
will
be
de
te
ct
ed.
This
thickne
ss
will
b
e
us
e
d
as
ref
e
ren
ce
to dete
ct
the sig
n of t
he prese
nce
of e
dem
a
m
acula on O
CT i
m
age.
ACKN
OWLE
DGE
MENTS
We
a
re
in
de
bted
to
the
e
xpe
rts
w
ho
hav
e
con
t
rib
uted
t
owar
ds
de
velo
pm
ent
of
the
te
m
pla
te
.
The
auth
or
s
wo
uld
li
ke
to ac
know
le
dg
e t
o
Guna
da
rm
a U
niv
er
sit
y.
REFERE
NCE
S
[1]
Bowling,
Kansk
is's
Cli
nic
al
Ophthalmology
A
Syste
mati
c
Approach,
Ei
ghth
Ed
it
ion,
ISBN
:
978
-
0
-
7020
-
5572
-
0
,
El
sevi
er,
S
y
dn
e
y,
New South
W
a
le
s,
Aus
trali
a. (
2
016)
[2]
Vaugha
n,
As
bury
,
Riord
an Eva.
Gene
ral Opht
hal
mology
.
Ed
ition
17.
London
:
Mc
Graw
-
Hill
.
2007
.
[3]
Raz
i
eh
Akhav
a
n,
Kari
m
Fa
ez,
“
A
Novel
Ret
inal
B
lood
Vess
el
Segm
ent
a
ti
on
Algor
i
thm
using
Fuz
z
y
Segm
ent
at
ion
”
,
I
nte
rnational
Jou
rnal
of El
e
ct
ri
ca
l
and
Comput
er
Engi
ne
ering
(
IJ
ECE
)
,
Vol.
4,
No,
2016
.
[4]
K.
Ga
y
a
thri,
D.
Narm
adha
,
K.
T
hil
ag
ava
th
i,
K.
Pavit
hr
a,
M.
Prad
ee
pa
,
“
Ret
inal
Blood
Vess
el
s
Ext
rac
t
ion
Based
o
n
Curvel
e
t
Tra
nsf
orm
and
b
y
Com
bini
ng
Botha
t
and
Topha
t
Mor
pholog
y
”
,
In
te
rn
ati
onal
Journal
of
El
e
ct
ri
cal
an
d
Computer
Engi
n
ee
ring (
IJE
C
E)
,
Vol.
4
,
No.
3,
20
14.
[5]
Li
ese
g
ang,
Skut
a,
Cant
or
.
Retin
a
and
Vi
treous.
Basic
and
Cli
ni
c
al
Cours
e.
Sect
i
on
12.
San
Fransisco,
Cal
iforn
ia:
Am
eri
ca
n
Ac
adem
y
of
Ophth
al
m
olog
y
.
2004
.
[6]
Sus
et
ia
ningtias,
D.T
.
,
Made
nd
a,
S.,
R
ahay
u
,
D.
A.,
Rodi
ah,
“
Ret
ina
l
m
ic
roa
n
eur
s
y
m
det
e
cti
on
u
sing
m
axi
m
all
y
stabl
e
externa
l
reg
ion
al
gori
th
m
”
,
Inte
rnat
ion
al
Journal
on
Adv
anc
ed
Sc
ie
n
ce
,
Engi
n
ee
ring
and
Informati
on
Technol
ogy
,
vo
l 6 No.
5
,
2016
.
[7]
Gianc
ard
,
Karno
ws
ki
,
Grisan
,
Fa
var
o
,
Rugg
eri
,
a
nd
Chaum.
“
Te
x
ture
l
ess
Mac
ula
Sw
el
li
ng
Det
ec
t
i
on
w
ith
Multi
p
le
Ret
inal
Fundus
I
m
age
s
”
.
I
EE
E
T
rans
act
ions
o
n
B
iomedi
ca
l
Eng
in
ee
ring
,
Vol.
58,
No.
3.
(
2011)
[8]
Fow
jiy
a
,
Karna
n,
Sivakumar.
“
An
Autom
at
ic
Dete
c
ti
on
and
A
ss
ess
m
ent
of
Diabe
t
ic
Mac
u
la
r
Ede
m
a
Along
with
Fovea
Dete
c
ti
on
from
Color
Ret
ina
l
Im
age
s
”
,
Int
ern
ati
onal
Journ
al
of
Computer
Tr
ends
and
Technol
ogy
(
IJCTT
)
.
Volum
e
4
Iss
ue 4.
2013
.
[9]
Anindit
a
Sept
ianirini,
Ham
dani
,
Dy
n
a
Marisa
Khair
ina,
“
The
Contour
Ext
ra
ction
of
Cup
in
F
undus
Im
age
s
f
or
Glauc
om
a
De
te
c
ti
on
”
,
Int
ernati
o
nal
Journal
of
Elec
tri
cal
and
Co
mputer
Engi
n
ee
r
ing
(
IJE
CE)
,
Vo
l.
6
,
No
.
6
,
2016
.
[10]
Deva
shree
.
Zi
nj
ard
e,
Mohota
.
“
Com
pute
r
Aided
Diagnosis
of
Mac
ula
r
Ed
ema
Us
ing
Color
F
undus
Im
age
s:
A
Revi
ew
”
.
In
t. J
o
urnal
of Engi
ne
e
ring R
ese
arch
a
nd
Applications
.
ISS
N:
2248
-
9622,
Vol
.
4
,
Iss
ue
3
.
2014
.
Evaluation Warning : The document was created with Spire.PDF for Python.