I
nte
rna
t
io
na
l J
o
urna
l o
f
Adv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
1
5
,
No
.
3
,
Sep
tem
b
er
20
2
6
,
p
p
.
1
1
1
2
~
1
1
2
2
I
SS
N:
2252
-
8
8
1
4
,
DOI
:
1
0
.
1
1
5
9
1
/ijaas
.
v
1
5
.
i
3
.
pp
1
1
1
2
-
1
1
2
2
1112
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
a
s
.
ia
esco
r
e.
co
m
Les
io
n
-
sem
a
ntic
t
o
ken guid
ed
trans
former f
usio
n f
o
r
sta
g
e
-
a
da
ptive dia
betic
retinopa
thy g
ra
di
ng
and screening
M
ura
li G
ujj
ula
,
K
um
a
r
Na
r
a
y
a
na
n
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
r
i
n
g
,
S
c
h
o
o
l
o
f
E
n
g
i
n
e
e
r
i
n
g
,
V
e
l
s I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
,
T
e
c
h
n
o
l
o
g
y
a
n
d
A
d
v
a
n
c
e
d
S
t
u
d
i
e
s,
C
h
e
n
n
a
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
Ma
y
6
,
2
0
2
5
R
ev
is
ed
Au
g
2
,
2
0
2
6
Acc
ep
ted
Au
g
1
3
,
2
0
2
6
Dia
b
e
ti
c
re
ti
n
o
p
a
t
h
y
(DR)
sc
re
e
n
in
g
fr
o
m
re
ti
n
a
l
f
u
n
d
u
s
ima
g
e
s
is
stil
l
d
iffi
c
u
lt
b
e
c
a
u
se
o
f
t
h
e
h
e
tero
g
e
n
e
it
y
o
f
les
io
n
s,
sig
n
ifi
c
a
n
t
c
las
s
i
m
b
a
lan
c
e
,
a
n
d
lo
w
re
li
a
b
il
i
ty
o
f
e
x
ist
in
g
d
e
e
p
lea
rn
i
n
g
a
l
g
o
rit
h
m
s
o
win
g
to
t
h
e
ir
g
e
n
e
ric
fe
a
tu
re
p
o
o
l
in
g
a
n
d
p
a
ss
i
v
e
fu
sio
n
m
e
c
h
a
n
ism
s.
To
o
v
e
rc
o
m
e
su
c
h
li
m
it
a
ti
o
n
s,
t
h
is
st
u
d
y
p
re
se
n
ts
a
h
y
b
ri
d
m
o
d
e
l
c
a
ll
e
d
Ha
n
d
Eff
iT
ra
n
s
-
DR,
wh
ich
c
o
m
b
i
n
e
s
d
e
e
p
re
ti
n
a
l
re
p
re
se
n
tatio
n
s
wit
h
h
a
n
d
c
ra
fted
les
io
n
d
e
sc
rip
to
rs
v
ia
t
h
e
d
irec
ted
tran
s
fo
rm
e
r
-
b
a
se
d
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
.
F
irst,
a
les
io
n
-
p
r
io
r
-
g
u
i
d
e
d
t
o
k
e
n
s
q
u
e
e
z
e
r
(LP
G
-
TS
)
m
o
d
u
le
is
p
ro
p
o
se
d
in
th
e
wo
rk
,
wh
ich
is
c
a
p
a
b
le
o
f
c
r
e
a
ti
n
g
les
io
n
-
sp
e
c
ifi
c
v
isu
a
l
to
k
e
n
s
v
ia
in
teg
ra
ti
n
g
p
rio
r
k
n
o
wle
d
g
e
a
b
o
u
t
p
a
th
o
lo
g
y
i
n
to
th
e
to
k
e
n
iza
ti
o
n
p
ro
c
e
d
u
re
.
M
o
re
o
v
e
r,
a
d
irec
ted
tran
sfo
rm
e
r
-
b
a
se
d
fu
sio
n
(
T→V
)
stra
teg
y
a
ll
o
ws
fo
r
a
c
ti
v
e
ly
f
u
sin
g
t
h
e
h
a
n
d
c
ra
fted
l
e
sio
n
fe
a
tu
re
s
wit
h
v
isu
a
l
re
p
re
s
e
n
tatio
n
s.
Th
e
in
tro
d
u
c
e
d
fra
m
e
wo
rk
is
a
ss
e
ss
e
d
with
a
stra
ti
fied
fiv
e
-
fo
ld
c
ro
ss
-
v
a
li
d
a
ti
o
n
p
r
o
to
c
o
l.
Its
a
v
e
ra
g
e
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
su
r
p
a
ss
e
s
9
7
%
,
wh
i
le
th
e
m
a
c
ro
-
F
1
sc
o
re
is
h
i
g
h
e
r
t
h
a
n
0
.
9
7
a
n
d
Co
h
e
n
'
s
k
a
p
p
a
is
g
r
e
a
ter
th
a
n
0
.
9
5
,
in
d
ica
ti
n
g
g
o
o
d
a
g
re
e
m
e
n
t
d
e
sp
it
e
t
h
e
p
re
se
n
c
e
o
f
c
las
s
i
m
b
a
lan
c
e
.
Th
e
p
ro
b
a
b
il
i
ty
re
li
a
b
i
li
ty
is
m
e
a
su
re
d
in
term
s
o
f
e
x
p
e
c
ted
c
a
li
b
r
a
ti
o
n
e
rro
r
(ECE
)
a
n
d
Brier
sc
o
re
s
a
n
d
s
h
o
ws
we
ll
-
c
a
li
b
ra
ted
c
o
n
f
id
e
n
c
e
.
F
u
rth
e
r
v
a
li
d
a
ti
o
n
is
p
e
rfo
rm
e
d
o
n
t
wo
e
x
tern
a
l
d
a
tas
e
ts,
wh
e
re
c
o
n
siste
n
t
p
e
rfo
rm
a
n
c
e
with
m
i
n
o
r
d
e
c
li
n
e
i
s sh
o
wn
.
K
ey
w
o
r
d
s
:
Diab
etic
r
etin
o
p
ath
y
L
esio
n
-
awa
r
e
to
k
e
n
izatio
n
Pro
b
ab
ilit
y
ca
lib
r
atio
n
R
etin
al
f
u
n
d
u
s
im
ag
in
g
T
r
an
s
f
o
r
m
e
r
atten
tio
n
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
:
Mu
r
ali
Gu
jju
la
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
Sch
o
o
l o
f
E
n
g
in
ee
r
in
g
Vels I
n
s
titu
te
o
f
Scien
ce
,
T
ec
h
n
o
lo
g
y
an
d
Ad
v
a
n
ce
d
Stu
d
ie
s
C
h
en
n
ai,
I
n
d
ia
E
m
ail:
h
ar
is
.
ee
e@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Diab
etic
r
etin
o
p
ath
y
(
DR
)
is
a
p
r
o
g
r
ess
iv
e
m
icr
o
v
ascu
lar
co
m
p
licatio
n
o
f
d
iab
etes
m
ellitu
s
an
d
r
em
ain
s
o
n
e
o
f
th
e
lea
d
in
g
ca
u
s
es
o
f
ir
r
e
v
er
s
ib
le
v
is
io
n
im
p
air
m
en
t
wo
r
ld
wid
e.
T
h
e
d
is
ea
s
e
m
an
if
ests
th
r
o
u
g
h
r
etin
al
lesi
o
n
s
s
u
ch
a
s
m
icr
o
an
eu
r
y
s
m
s
,
h
em
o
r
r
h
ag
es,
h
ar
d
e
x
u
d
ates,
an
d
n
e
o
v
as
cu
lar
izatio
n
,
wh
ich
o
f
ten
ap
p
ea
r
at
ea
r
l
y
s
tag
es
with
s
u
b
tle
v
is
u
al
co
n
tr
ast
an
d
lo
ca
lized
s
p
atial
d
is
tr
i
b
u
tio
n
.
L
a
r
g
e
-
s
ca
le
s
cr
ee
n
in
g
u
s
in
g
f
u
n
d
u
s
p
h
o
to
g
r
ap
h
y
is
th
e
r
ef
o
r
e
ess
en
tial
f
o
r
e
ar
ly
d
iag
n
o
s
is
an
d
tim
ely
in
ter
v
en
tio
n
,
y
e
t
m
an
u
al
s
cr
ee
n
in
g
is
lim
ited
b
y
s
p
ec
ialis
t
av
ailab
ilit
y
,
i
n
ter
-
o
b
s
er
v
er
v
ar
iab
ilit
y
,
a
n
d
in
cr
ea
s
in
g
p
atien
t
b
u
r
d
e
n
.
As
a
r
esu
lt,
au
to
m
ate
d
DR
d
etec
tio
n
u
s
in
g
ar
tific
ia
l
in
tellig
en
ce
h
as
b
ec
o
m
e
a
n
a
ctiv
e
r
esear
ch
a
r
ea
,
aim
in
g
to
s
u
p
p
o
r
t
o
p
h
th
alm
o
l
o
g
is
ts
b
y
p
r
o
v
id
in
g
f
ast,
co
n
s
i
s
ten
t,
an
d
s
ca
lab
le
s
cr
ee
n
in
g
s
o
lu
tio
n
s
[
1
]
,
[
2
]
.
Dee
p
lear
n
in
g
,
p
a
r
ticu
lar
ly
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs),
h
as
d
em
o
n
s
tr
at
ed
s
tr
o
n
g
p
er
f
o
r
m
an
ce
in
DR
d
etec
tio
n
b
y
lear
n
in
g
h
ier
ar
c
h
ical
r
e
p
r
esen
tatio
n
s
d
ir
ec
tly
f
r
o
m
r
eti
n
al
im
ag
es
[
3
]
,
[
4
]
.
C
NN
-
b
ased
s
y
s
tem
s
h
av
e
ac
h
iev
ed
h
ig
h
ac
cu
r
ac
y
o
n
p
u
b
lic
b
en
c
h
m
ar
k
s
an
d
ar
e
in
cr
ea
s
in
g
ly
e
x
p
l
o
r
e
d
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Les
io
n
-
s
ema
n
tic
to
ke
n
g
u
id
e
d
tr
a
n
s
fo
r
mer fu
s
io
n
fo
r
s
ta
g
e
-
a
d
a
p
tive
d
ia
b
etic
r
etin
o
p
a
th
y
…
(
Mu
r
a
li Gu
jju
la
)
1113
r
ea
l
-
wo
r
ld
d
e
p
lo
y
m
e
n
t
[
5
]
.
Ho
wev
er
,
d
esp
ite
th
eir
s
u
cc
ess
,
C
NN
-
o
n
ly
ap
p
r
o
ac
h
es
o
f
ten
f
u
n
ctio
n
as
b
lack
-
b
o
x
m
o
d
els,
o
f
f
er
i
n
g
li
m
ited
in
ter
p
r
etab
ilit
y
an
d
r
ed
u
ce
d
r
o
b
u
s
tn
ess
wh
en
ex
p
o
s
ed
to
v
ar
iatio
n
s
in
a
c
q
u
i
s
it
i
o
n
d
e
v
i
c
es
,
il
l
u
m
i
n
at
io
n
c
o
n
d
i
t
i
o
n
s
,
o
r
p
a
ti
e
n
t
d
e
m
o
g
r
a
p
h
i
c
s
[
6
]
,
[
7
]
.
T
h
e
s
e
l
i
m
i
ta
t
i
o
n
s
a
r
e
c
r
it
i
c
a
l
i
n
m
e
d
i
c
a
l
s
c
r
e
e
n
i
n
g
,
w
h
e
r
e
m
o
d
e
l
r
e
l
i
a
b
i
l
i
t
y
a
n
d
t
r
a
n
s
p
a
r
e
n
c
y
ar
e
a
s
i
m
p
o
r
t
a
n
t
a
s
r
aw
c
l
as
s
i
f
i
ca
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
.
T
o
o
v
er
co
m
e
s
h
o
r
tco
m
in
g
s
o
f
p
u
r
ely
d
ee
p
m
o
d
els,
r
ec
en
t
s
tu
d
ies
h
av
e
ex
p
lo
r
e
d
h
y
b
r
id
f
r
am
ewo
r
k
s
th
at
co
m
b
in
e
d
e
e
p
f
ea
tu
r
es
with
h
an
d
cr
af
ted
r
etin
al
d
escr
ip
to
r
s
an
d
atten
tio
n
m
e
ch
an
is
m
s
[
8
]
,
[
9
]
.
Han
d
cr
af
ted
f
ea
tu
r
es,
s
u
ch
a
s
tex
tu
r
e
s
tatis
tics
an
d
lesi
o
n
-
lev
el
d
escr
ip
to
r
s
,
e
n
co
d
e
clin
ically
m
ea
n
in
g
f
u
l
in
f
o
r
m
atio
n
th
at
co
m
p
lem
en
t
s
d
ee
p
r
ep
r
esen
tatio
n
s
an
d
e
n
h
an
ce
s
in
ter
p
r
etab
ilit
y
[
1
0
]
.
At
th
e
s
am
e
tim
e,
tr
an
s
f
o
r
m
er
-
b
ased
ar
ch
itectu
r
es
an
d
v
is
io
n
tr
an
s
f
o
r
m
er
s
h
a
v
e
b
ee
n
in
t
r
o
d
u
ce
d
in
to
r
etin
a
l
im
ag
e
an
aly
s
is
to
m
o
d
el
lo
n
g
-
r
an
g
e
d
ep
e
n
d
en
c
ies
an
d
g
lo
b
al
co
n
tex
t
th
r
o
u
g
h
s
elf
-
atten
tio
n
m
ec
h
a
n
is
m
s
[
1
1
]
,
[
1
2
]
.
T
h
ese
ap
p
r
o
ac
h
es
h
av
e
s
h
o
wn
p
r
o
m
is
e
in
ca
p
tu
r
in
g
s
p
atial
r
elatio
n
s
h
ip
s
b
etwe
en
lesi
o
n
s
an
d
an
ato
m
ical
s
tr
u
ctu
r
es
th
at
ar
e
d
if
f
icu
lt t
o
m
o
d
el
u
s
in
g
lo
ca
l c
o
n
v
o
lu
tio
n
alo
n
e
[
1
3
]
.
Mo
r
e
r
ec
en
t
wo
r
k
s
h
av
e
p
r
o
p
o
s
e
d
C
NN
-
tr
a
n
s
f
o
r
m
er
f
u
s
io
n
m
o
d
els
f
o
r
DR
d
ete
cti
o
n
an
d
g
r
a
d
i
n
g
,
r
e
p
o
r
ti
n
g
im
p
r
o
v
e
d
d
is
c
r
i
m
i
n
a
tio
n
p
er
f
o
r
m
a
n
c
e
co
m
p
a
r
e
d
to
s
ta
n
d
al
o
n
e
C
N
Ns
[
1
4
]
,
[
1
5
]
.
Nev
er
t
h
el
ess
,
m
o
s
t
ex
is
ti
n
g
f
u
s
i
o
n
s
t
r
a
te
g
ies
r
e
ly
o
n
s
i
m
p
le
f
e
at
u
r
e
c
o
n
ca
te
n
at
i
o
n
o
r
s
y
m
m
et
r
ic
a
tte
n
t
io
n
,
wh
ich
t
r
ea
ts
a
ll
s
p
ati
al
r
e
g
i
o
n
s
u
n
i
f
o
r
m
l
y
a
n
d
f
a
ils
t
o
ex
p
li
cit
ly
em
p
h
asi
ze
l
esi
o
n
-
d
o
m
i
n
a
n
t
a
r
e
as
[
1
6
]
,
[
1
7
]
.
F
u
r
t
h
e
r
m
o
r
e
,
h
an
d
cr
af
te
d
f
e
at
u
r
es,
w
h
en
in
c
o
r
p
o
r
at
ed
,
ar
e
o
f
t
en
a
p
p
e
n
d
ed
o
n
l
y
a
t
th
e
cl
ass
if
ie
r
s
t
ag
e,
lim
iti
n
g
th
e
ir
in
f
l
u
en
ce
o
n
in
t
er
m
e
d
i
at
e
r
e
p
r
es
en
tati
o
n
le
ar
n
i
n
g
[
1
8
]
.
An
o
t
h
e
r
c
r
it
ic
al
l
im
it
ati
o
n
is
th
at
m
a
n
y
s
t
u
d
ies
f
o
c
u
s
p
r
im
ar
il
y
o
n
ac
c
u
r
a
cy
o
r
a
r
e
a
u
n
d
er
th
e
r
e
ce
i
v
e
r
o
p
e
r
at
in
g
ch
ar
ac
te
r
is
ti
c
cu
r
v
e
(
AUROC
)
,
wh
ile
o
v
e
r
lo
o
k
i
n
g
p
r
o
b
ab
ilit
y
ca
li
b
r
at
io
n
a
n
d
co
n
f
i
d
e
n
ce
r
eli
ab
i
lit
y
—
f
a
ct
o
r
s
t
h
a
t
ar
e
cr
u
c
ia
l
f
o
r
cli
n
i
ca
l
d
e
cisi
o
n
-
m
a
k
i
n
g
[
1
9
]
,
[
2
0
]
.
Desp
ite
s
ig
n
if
ican
t
p
r
o
g
r
ess
,
s
ev
er
al
g
ap
s
r
em
ain
i
n
cu
r
r
en
t
DR
d
etec
tio
n
m
eth
o
d
o
l
o
g
ies.
First,
ex
is
tin
g
m
o
d
els
r
a
r
ely
in
te
g
r
ate
ex
p
licit
lesi
o
n
p
r
io
r
s
in
to
th
e
to
k
e
n
izatio
n
o
r
atten
tio
n
p
r
o
ce
s
s
,
lead
in
g
to
d
ilu
ted
lesi
o
n
in
f
o
r
m
atio
n
wh
en
f
ea
tu
r
es
ar
e
g
lo
b
ally
p
o
o
led
o
r
u
n
if
o
r
m
ly
p
ar
titi
o
n
ed
.
Seco
n
d
,
t
h
e
in
ter
ac
tio
n
b
etwe
en
h
an
d
c
r
af
ted
lesi
o
n
d
escr
ip
to
r
s
an
d
d
ee
p
v
is
u
al
r
ep
r
esen
tatio
n
s
is
o
f
ten
wea
k
o
r
p
ass
iv
e,
p
r
ev
en
tin
g
clin
ically
m
ea
n
in
g
f
u
l
g
u
id
an
ce
d
u
r
in
g
f
ea
tu
r
e
f
u
s
io
n
.
T
h
ir
d
,
tr
an
s
f
o
r
m
e
r
-
b
ased
f
u
s
io
n
m
ec
h
an
is
m
s
lack
d
ir
ec
tio
n
ality
th
at
r
ef
lects
clin
ica
l
r
ea
s
o
n
in
g
,
wh
er
e
lesi
o
n
s
t
atis
tics
s
h
o
u
ld
in
f
o
r
m
v
is
u
al
in
ter
p
r
etatio
n
r
at
h
er
th
an
th
e
r
ev
er
s
e
[
2
1
]
,
[
2
2
]
.
Fin
ally
,
m
an
y
r
ep
o
r
ted
s
y
s
te
m
s
ar
e
ev
alu
ate
d
o
n
im
b
ala
n
ce
d
d
atasets
u
s
in
g
ac
cu
r
ac
y
-
ce
n
tr
ic
m
etr
ics,
with
o
u
t
s
u
f
f
icien
t
a
n
aly
s
is
o
f
c
lass
-
wi
s
e
b
eh
av
io
r
,
r
o
b
u
s
tn
ess
ac
r
o
s
s
f
o
ld
s
,
o
r
p
r
o
b
a
b
ilit
y
ca
lib
r
atio
n
,
w
h
ich
lim
its
th
eir
s
u
itab
ilit
y
f
o
r
r
ea
l
-
wo
r
ld
s
cr
ee
n
in
g
s
ce
n
ar
io
s
[
2
3
]
,
[
2
4
]
.
T
h
e
p
r
o
b
le
m
ad
d
r
ess
ed
in
th
is
wo
r
k
is
th
er
ef
o
r
e
th
e
d
ev
el
o
p
m
en
t
o
f
a
DR
d
etec
tio
n
f
r
am
e
wo
r
k
th
at
i)
ex
p
licitly
em
p
h
asizes
les
i
o
n
-
r
elev
a
n
t
r
eg
io
n
s
d
u
r
i
n
g
r
ep
r
esen
tatio
n
lear
n
in
g
;
ii)
en
ab
les
ac
tiv
e
an
d
in
ter
p
r
etab
le
i
n
ter
ac
tio
n
b
etw
ee
n
h
a
n
d
cr
af
te
d
lesi
o
n
f
ea
tu
r
e
s
an
d
d
ee
p
v
is
u
al
to
k
en
s
;
an
d
iii)
r
em
ai
n
s
r
o
b
u
s
t
an
d
r
eliab
le
u
n
d
er
class
im
b
alan
ce
an
d
d
ataset
v
ar
iab
ilit
y
.
T
o
ad
d
r
ess
th
e
id
e
n
tifie
d
g
a
p
s
,
th
is
wo
r
k
p
r
o
p
o
s
es
Han
d
E
f
f
iTr
a
n
s
-
DR
,
a
lesi
o
n
-
awa
r
e
h
y
b
r
i
d
d
ee
p
lear
n
in
g
f
r
a
m
ewo
r
k
f
o
r
au
t
o
m
ated
DR
d
et
ec
tio
n
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
in
tr
o
d
u
ce
s
two
k
ey
in
n
o
v
atio
n
s
.
First,
a
lesi
o
n
-
p
r
io
r
-
g
u
id
ed
to
k
e
n
s
q
u
ee
ze
r
(
L
PG
-
T
S)
g
en
er
ates
lesi
o
n
-
ce
n
tr
ic
v
is
u
al
to
k
en
s
b
y
ex
p
licitly
in
co
r
p
o
r
atin
g
lesi
o
n
p
r
io
r
m
ap
s
d
er
iv
e
d
f
r
o
m
r
etin
al
p
r
e
-
p
r
o
ce
s
s
in
g
.
T
h
is
en
s
u
r
es
th
at
to
k
en
r
ep
r
esen
tati
o
n
s
ar
e
s
p
atially
an
ch
o
r
ed
to
p
ath
o
lo
g
ical
r
e
g
io
n
s
r
ath
er
th
an
u
n
if
o
r
m
ly
s
am
p
le
d
b
ac
k
g
r
o
u
n
d
ar
ea
s
.
Seco
n
d
,
a
d
ir
ec
ted
tr
an
s
f
o
r
m
er
-
b
ased
f
u
s
io
n
m
ec
h
a
n
is
m
(
T
→V
)
is
em
p
lo
y
ed
,
wh
er
e
h
an
d
cr
af
ted
lesi
o
n
d
escr
ip
to
r
s
g
u
id
e
th
e
atten
tio
n
r
ef
in
em
e
n
t
o
f
v
is
u
al
to
k
en
s
,
en
ab
lin
g
clin
ically
m
ea
n
in
g
f
u
l
cr
o
s
s
-
m
o
d
al
r
ea
s
o
n
in
g
.
I
n
ad
d
itio
n
,
th
e
f
r
am
ewo
r
k
i
n
teg
r
ates
a
lesi
o
n
s
u
m
m
ar
y
to
k
en
t
o
p
r
eser
v
e
g
lo
b
al
r
etin
al
c
o
n
tex
t
a
n
d
e
m
p
lo
y
s
ca
lib
r
atio
n
-
awa
r
e
e
v
alu
atio
n
to
ass
ess
th
e
r
eliab
ilit
y
o
f
p
r
ed
icted
p
r
o
b
ab
ilit
ies.
T
h
e
o
v
er
all
ar
ch
itectu
r
e
is
d
esig
n
ed
to
b
alan
ce
d
is
cr
im
in
ativ
e
p
er
f
o
r
m
an
ce
,
in
ter
p
r
etab
ilit
y
,
an
d
r
o
b
u
s
tn
ess
,
alig
n
in
g
with
th
e
p
r
ac
tical
r
eq
u
i
r
em
en
ts
o
f
DR
s
cr
ee
n
in
g
s
y
s
tem
s
.
2.
M
E
T
H
O
D
2
.
1
.
Da
t
a
s
et
cha
ra
c
t
er
is
t
ics a
nd
cla
s
s
im
ba
la
nce
co
ns
ide
ra
t
io
n
Pu
b
lic
DR
d
atasets
ar
e
in
h
er
en
tly
class
-
im
b
alan
ce
d
,
wi
th
n
o
r
m
al
o
r
m
ild
ca
s
es
s
ig
n
if
ican
tly
o
u
tn
u
m
b
er
in
g
m
o
d
er
ate
-
to
-
s
ev
er
e
d
is
ea
s
e
s
am
p
les.
T
h
is
i
m
b
alan
ce
p
o
s
es
a
m
ajo
r
ch
a
llen
g
e,
as
m
o
d
els
tr
ain
ed
with
o
u
t
ap
p
r
o
p
r
iate
c
o
n
s
id
er
atio
n
m
a
y
ac
h
iev
e
h
ig
h
ac
cu
r
a
cy
w
h
ile
u
n
d
e
r
p
er
f
o
r
m
in
g
o
n
clin
ically
cr
itical
m
in
o
r
ity
class
es.
I
n
t
h
is
s
tu
d
y
,
s
tr
atif
ied
f
iv
e
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
is
em
p
lo
y
e
d
to
p
r
eser
v
e
class
d
is
tr
ib
u
tio
n
ac
r
o
s
s
f
o
l
d
s
,
en
s
u
r
in
g
f
air
a
n
d
u
n
b
iased
ev
alu
ati
o
n
.
Mo
r
eo
v
er
,
p
e
r
f
o
r
m
an
ce
is
ass
ess
ed
u
s
in
g
m
ac
r
o
-
av
er
a
g
ed
m
etr
ics,
C
o
h
en
’
s
k
a
p
p
a,
an
d
ca
lib
r
atio
n
m
ea
s
u
r
es,
wh
ich
p
r
o
v
id
e
a
m
o
r
e
in
f
o
r
m
ativ
e
ass
ess
m
en
t
u
n
d
er
i
m
b
alan
ce
co
m
p
a
r
ed
to
ac
cu
r
ac
y
alo
n
e
[
2
5
]
.
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
is
ev
alu
ated
o
n
a
p
r
im
a
r
y
DR
d
ataset
an
d
f
u
r
th
er
v
alid
ated
o
n
in
d
ep
en
d
en
t
d
atasets
to
ass
es
s
g
en
er
aliza
tio
n
ac
r
o
s
s
d
if
f
er
en
t
d
ata
d
is
tr
ib
u
tio
n
s
.
B
y
co
m
b
in
in
g
lesi
o
n
-
awa
r
e
m
o
d
elin
g
,
d
ir
ec
te
d
f
u
s
io
n
,
an
d
im
b
ala
n
ce
-
s
en
s
itiv
e
ev
alu
atio
n
,
th
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
aim
s
to
d
eliv
er
a
r
el
iab
le
an
d
clin
ically
m
ea
n
in
g
f
u
l so
lu
tio
n
f
o
r
a
u
to
m
ated
DR
s
cr
ee
n
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J
Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
3
,
Sep
tem
b
er
2
0
2
6
:
1
1
1
2
-
1
1
2
2
1114
2
.
2
.
O
v
er
v
iew
o
f
t
he
pro
po
s
ed
f
r
a
m
ewo
r
k
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
,
r
e
f
er
r
ed
t
o
as
Han
d
E
f
f
iTr
an
s
-
DR
,
is
a
lesi
o
n
-
awa
r
e
h
y
b
r
id
lear
n
in
g
ar
ch
itectu
r
e
d
esig
n
ed
f
o
r
ac
c
u
r
ate
an
d
r
elia
b
le
DR
d
etec
t
io
n
f
r
o
m
r
etin
al
f
u
n
d
u
s
im
a
g
es.
T
h
e
co
m
p
lete
p
r
o
ce
s
s
in
g
p
ip
elin
e
is
illu
s
tr
ated
in
Fig
u
r
e
1
,
wh
ich
p
r
esen
ts
th
e
en
d
-
to
-
e
n
d
f
lo
w
f
r
o
m
i
m
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
to
ca
lib
r
ated
DR
p
r
ed
ictio
n
.
T
h
e
ce
n
tr
al
d
esig
n
p
h
ilo
s
o
p
h
y
o
f
th
e
f
r
am
ewo
r
k
is
to
en
s
u
r
e
th
at
p
ath
o
lo
g
ical
lesi
o
n
in
f
o
r
m
atio
n
is
ex
p
licitl
y
em
b
e
d
d
ed
in
to
b
o
th
f
ea
tu
r
e
r
ep
r
esen
tatio
n
an
d
f
u
s
io
n
s
tag
es,
r
ath
er
th
an
b
ein
g
im
p
licitly
lear
n
ed
th
r
o
u
g
h
d
e
ep
n
etwo
r
k
s
alo
n
e.
T
o
ac
h
ie
v
e
th
is
,
t
h
e
f
r
am
ewo
r
k
i
n
teg
r
ates
lesi
o
n
-
ce
n
tr
ic
v
is
u
al
to
k
en
izatio
n
,
h
a
n
d
cr
af
t
ed
r
etin
al
d
escr
ip
to
r
s
,
an
d
a
d
ir
ec
ted
tr
an
s
f
o
r
m
er
-
b
ased
f
u
s
i
o
n
m
ec
h
an
is
m
.
As
s
h
o
wn
in
Fig
u
r
e
1
,
th
e
m
et
h
o
d
o
lo
g
y
co
n
s
is
ts
o
f
f
iv
e
s
eq
u
en
tial
s
tag
es:
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
lesi
o
n
en
h
an
ce
m
e
n
t,
p
a
r
allel
d
ee
p
an
d
h
an
d
cr
af
ted
f
ea
tu
r
e
ex
tr
ac
ti
o
n
,
lesi
o
n
-
p
r
io
r
-
g
u
id
ed
to
k
en
g
en
er
atio
n
,
d
ir
ec
ted
tr
an
s
f
o
r
m
er
f
u
s
io
n
,
an
d
ca
lib
r
ated
class
if
icatio
n
.
T
h
ese
s
tag
es
ar
e
t
ig
h
tly
co
u
p
led
to
m
ain
tain
co
n
s
is
ten
cy
b
etwe
en
p
ath
o
l
o
g
ical
ev
id
e
n
c
e
an
d
f
in
al
d
ec
is
io
n
-
m
ak
i
n
g
.
Fig
u
r
e
1
.
Ov
e
r
all
Han
d
E
f
f
iTr
a
n
s
-
DR
f
r
am
ewo
r
k
ar
c
h
itectu
r
e
2
.
3
.
I
ma
g
e
pre
-
pro
ce
s
s
ing
a
nd
lesi
o
n e
n
ha
ncem
ent
E
ac
h
in
p
u
t
f
u
n
d
u
s
im
ag
e
is
f
ir
s
t
r
esized
to
a
f
i
x
ed
s
p
ati
al
r
eso
lu
tio
n
an
d
n
o
r
m
alize
d
to
r
ed
u
ce
illu
m
in
atio
n
v
ar
iatio
n
s
co
m
m
o
n
ly
o
b
s
er
v
ed
ac
r
o
s
s
s
cr
ee
n
in
g
d
e
v
ices.
C
o
n
tr
ast
en
h
an
ce
m
en
t
is
ap
p
lied
t
o
im
p
r
o
v
e
th
e
v
is
ib
ilit
y
o
f
s
u
b
tle
DR
m
an
if
estatio
n
s
s
u
ch
a
s
m
icr
o
-
an
e
u
r
y
s
m
s
an
d
h
em
o
r
r
h
a
g
es
.
No
is
e
s
u
p
p
r
ess
io
n
an
d
m
o
r
p
h
o
lo
g
ic
al
r
ef
in
em
en
t
a
r
e
s
u
b
s
eq
u
e
n
tly
em
p
lo
y
ed
to
p
r
eser
v
e
lesi
o
n
b
o
u
n
d
ar
ies
wh
ile
s
u
p
p
r
ess
in
g
b
ac
k
g
r
o
u
n
d
ar
tef
ac
ts
.
T
h
is
s
tag
e
p
r
o
d
u
ce
s
an
en
h
an
ce
d
r
etin
al
im
ag
e
th
at
s
u
p
p
o
r
ts
b
o
th
d
ee
p
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
lesi
o
n
lo
ca
lizatio
n
.
T
h
e
lesi
o
n
e
n
h
a
n
ce
m
en
t
p
r
o
ce
s
s
also
g
en
er
at
es
a
b
in
ar
y
lesi
o
n
m
ask
,
wh
ich
is
later
tr
an
s
f
o
r
m
ed
in
to
a
lesi
o
n
p
r
i
o
r
m
ap
.
I
m
p
o
r
ta
n
tly
,
th
is
lesi
o
n
p
r
i
o
r
is
n
o
t
tr
ea
ted
as
an
au
x
iliar
y
o
u
tp
u
t;
in
s
tead
,
it
p
lay
s
a
d
ir
ec
t
r
o
l
e
in
g
u
id
in
g
r
ep
r
esen
tatio
n
lear
n
in
g
,
f
o
r
m
i
n
g
a
f
o
u
n
d
atio
n
al
co
m
p
o
n
en
t o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
.
2
.
4
.
Dee
p f
e
a
t
ure
ex
t
r
a
ct
io
n
a
nd
lesi
o
n prio
r
m
o
dellin
g
Dee
p
v
is
u
al
f
ea
tu
r
es
ar
e
e
x
tr
ac
ted
u
s
in
g
a
p
r
etr
ain
ed
E
f
f
i
cien
tNet
-
B
0
b
ac
k
b
o
n
e,
wh
ic
h
r
em
ain
s
f
r
o
z
en
d
u
r
i
n
g
tr
ai
n
in
g
to
p
r
e
s
er
v
e
g
en
er
aliza
b
le
r
etin
al
r
e
p
r
esen
tatio
n
s
.
Fo
r
a
g
iv
e
n
p
r
e
-
p
r
o
ce
s
s
ed
f
u
n
d
u
s
im
ag
e
,
th
e
b
ac
k
b
o
n
e
p
r
o
d
u
ce
s
a
s
p
atial
f
ea
tu
r
e
m
ap
,
as d
ef
in
ed
in
(
1
)
.
=
(
)
∈
ℝ
×
×
(
1
)
W
h
er
e
an
d
d
en
o
te
t
h
e
s
p
atial
d
im
en
s
io
n
s
o
f
th
e
f
ea
tu
r
e
m
ap
,
an
d
r
ep
r
esen
ts
th
e
c
h
an
n
el
d
ep
th
.
T
h
is
f
ea
tu
r
e
m
ap
r
etain
s
b
o
th
s
em
a
n
tic
an
d
s
p
atial
in
f
o
r
m
atio
n
r
e
lev
an
t
to
DR
p
ath
o
lo
g
y
.
I
n
p
a
r
allel,
th
e
en
h
an
ce
d
lesi
o
n
m
ask
is
r
esized
to
m
at
ch
th
e
s
p
atial
r
eso
lu
tio
n
o
f
,
f
o
r
m
in
g
a
lesi
o
n
p
r
io
r
m
ap
∈
ℝ
×
×
1
.
T
h
e
co
ex
is
ten
ce
o
f
an
d
en
ab
les
th
e
f
r
am
ewo
r
k
t
o
r
ea
s
o
n
jo
i
n
tly
a
b
o
u
t
lear
n
ed
r
etin
al
s
em
an
tics
an
d
ex
p
licitly
lo
ca
lized
lesi
o
n
r
eg
io
n
s
,
wh
ich
is
ess
en
tial f
o
r
p
ath
o
lo
g
y
-
aw
ar
e
m
o
d
ellin
g
.
2
.
5
.
L
esio
n
-
prio
r
-
g
ui
ded t
o
ken sq
ueez
er
T
o
tr
an
s
f
o
r
m
s
p
atial
f
ea
tu
r
e
m
ap
s
in
t
o
a
co
m
p
ac
t
y
et
p
ath
o
l
o
g
y
-
f
o
cu
s
ed
r
ep
r
esen
tatio
n
,
th
e
f
r
am
ewo
r
k
em
p
lo
y
s
L
PG
-
T
S,
as
illu
s
tr
ated
in
Fig
u
r
e
2
.
Un
lik
e
c
o
n
v
e
n
tio
n
al
p
o
o
l
in
g
o
r
p
atch
-
b
ased
to
k
en
izatio
n
,
L
PG
-
T
S
g
en
er
a
tes
les
io
n
-
ce
n
tr
ic
v
is
u
al
to
k
en
s
b
y
ad
ap
tiv
ely
f
u
s
in
g
b
ac
k
b
o
n
e
s
alien
cy
an
d
lesi
o
n
p
r
io
r
in
f
o
r
m
atio
n
.
T
h
e
f
u
s
io
n
p
r
o
ce
s
s
is
g
o
v
e
r
n
ed
b
y
s
et
o
f
lear
n
a
b
le
g
ates a
n
d
is
f
o
r
m
u
lated
as
in
(
2
)
.
=
(
)
⊙
(
)
+
(
1
−
(
)
)
⊙
(
)
(
2
)
W
h
er
e
(
⋅
)
an
d
(
⋅
)
ar
e
1
×
1
co
n
v
o
lu
tio
n
al
p
r
o
jectio
n
o
f
th
e
f
ea
t
u
r
e
m
ap
an
d
lesi
o
n
p
r
io
r
m
a
p
,
r
esp
ec
tiv
ely
;
d
en
o
tes
lear
n
ab
le
g
atin
g
p
ar
am
eter
s
;
an
d
(
⋅
)
r
ep
r
esen
ts
th
e
s
ig
m
o
id
ac
tiv
atio
n
.
T
h
is
f
o
r
m
u
latio
n
en
a
b
les
th
e
m
o
d
el
to
d
y
n
am
ically
b
ala
n
ce
lear
n
ed
v
is
u
al
s
alien
cy
an
d
ex
p
licit
lesi
o
n
ev
id
en
ce
,
d
ep
en
d
in
g
o
n
im
ag
e
q
u
ality
an
d
lesi
o
n
p
r
o
m
in
en
c
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Les
io
n
-
s
ema
n
tic
to
ke
n
g
u
id
e
d
tr
a
n
s
fo
r
mer fu
s
io
n
fo
r
s
ta
g
e
-
a
d
a
p
tive
d
ia
b
etic
r
etin
o
p
a
th
y
…
(
Mu
r
a
li Gu
jju
la
)
1115
Sp
atial
s
o
f
tm
ax
is
ap
p
lied
to
th
e
f
u
s
ed
s
co
r
e
m
ap
S
to
o
b
t
ain
n
o
r
m
alize
d
atten
tio
n
weig
h
ts
.
T
h
ese
weig
h
ts
ar
e
th
en
u
s
ed
to
a
g
g
r
eg
ate
s
p
atial
f
ea
tu
r
es
in
to
a
f
i
x
ed
n
u
m
b
er
o
f
lesi
o
n
-
ce
n
t
r
ic
to
k
en
s
.
T
h
is
p
r
o
ce
s
s
en
s
u
r
es
th
at
th
e
r
esu
ltin
g
to
k
e
n
s
eq
u
en
ce
co
n
ce
n
tr
ates
r
e
p
r
e
s
en
tatio
n
al
ca
p
ac
ity
o
n
clin
ica
lly
r
elev
an
t
r
eg
io
n
s
r
ath
er
th
an
u
n
if
o
r
m
ly
s
am
p
le
d
b
ac
k
g
r
o
u
n
d
ar
ea
s
.
Fig
u
r
e
2
.
Sch
em
atic
d
iag
r
am
o
f
th
e
L
PG
-
TS
2
.
6
.
H
a
nd
cr
a
f
t
ed
f
ea
t
ure
mo
dellin
g
a
nd
lesi
o
n
s
um
m
a
ry
t
o
k
en
I
n
p
ar
allel
with
d
ee
p
f
ea
t
u
r
e
p
r
o
ce
s
s
in
g
,
h
an
d
cr
a
f
ted
r
etin
al
d
escr
ip
to
r
s
ca
p
tu
r
i
n
g
tex
tu
r
e
ch
ar
ac
ter
is
tics
an
d
lesi
o
n
p
r
o
x
y
s
tatis
tics
ar
e
ex
tr
ac
ted
.
T
h
ese
d
escr
ip
to
r
s
en
c
o
d
e
clin
i
ca
lly
in
ter
p
r
etab
le
in
f
o
r
m
atio
n
s
u
ch
as
in
ten
s
ity
v
ar
iatio
n
,
s
tr
u
ctu
r
al
ir
r
eg
u
lar
it
y
,
an
d
lesi
o
n
co
m
p
o
n
en
t
d
is
tr
ib
u
tio
n
.
T
o
en
a
b
le
in
ter
ac
tio
n
with
tr
an
s
f
o
r
m
er
-
b
ased
atten
tio
n
,
th
e
h
an
d
cr
af
te
d
f
ea
tu
r
e
v
ec
t
o
r
is
p
r
o
jecte
d
i
n
to
th
e
s
am
e
laten
t
em
b
ed
d
in
g
s
p
ac
e
an
d
r
ep
licat
ed
to
f
o
r
m
a
s
eq
u
e
n
ce
o
f
h
an
d
cr
af
ted
to
k
e
n
s
,
as illu
s
tr
ated
in
Fig
u
r
e
3
.
Fig
u
r
e
3
.
L
esio
n
s
u
m
m
ar
y
to
k
en
an
d
h
an
d
c
r
af
ted
to
k
en
m
o
d
ellin
g
Ad
d
itio
n
ally
,
a
lesi
o
n
s
u
m
m
ar
y
to
k
e
n
is
g
en
er
ate
d
f
r
o
m
th
e
b
ac
k
b
o
n
e
f
ea
tu
r
e
m
ap
u
s
in
g
li
g
h
tweig
h
t
co
n
v
o
l
u
tio
n
,
f
o
llo
wed
b
y
g
l
o
b
al
av
er
ag
e
p
o
o
lin
g
an
d
l
in
ea
r
p
r
o
jectio
n
.
T
h
is
t
o
k
en
p
r
eser
v
es
g
lo
b
al
r
etin
al
co
n
tex
t
th
at
m
ay
n
o
t
b
e
f
u
lly
ca
p
tu
r
ed
b
y
lesi
o
n
-
c
en
tr
ic
to
k
en
s
alo
n
e.
T
o
g
eth
er
,
th
e
lesi
o
n
-
ce
n
tr
ic
to
k
en
s
,
lesi
o
n
s
u
m
m
ar
y
to
k
e
n
,
an
d
h
an
d
c
r
af
ted
to
k
e
n
s
f
o
r
m
a
u
n
if
ie
d
m
u
ltimo
d
al
t
o
k
en
s
et
s
u
itab
le
f
o
r
tr
an
s
f
o
r
m
er
-
b
ased
f
u
s
io
n
.
2
.
7
.
Dire
ct
ed
t
r
a
ns
f
o
rm
er
f
u
s
io
n
C
r
o
s
s
-
m
o
d
al
in
teg
r
atio
n
is
p
e
r
f
o
r
m
e
d
u
s
in
g
a
d
ir
ec
ted
tr
a
n
s
f
o
r
m
er
f
u
s
io
n
m
e
ch
an
is
m
,
as
s
h
o
wn
in
Fig
u
r
e
4
.
I
n
th
is
d
esig
n
,
v
is
u
al
to
k
en
s
ac
t
as
q
u
er
ies,
wh
ile
h
an
d
cr
a
f
ted
to
k
e
n
s
s
er
v
e
as
k
ey
s
an
d
v
alu
es.
T
h
is
d
ir
ec
tio
n
ality
en
s
u
r
es
t
h
at
h
an
d
cr
af
te
d
lesi
o
n
d
escr
ip
to
r
s
ac
tiv
ely
g
u
i
d
e
th
e
r
ef
in
em
en
t
o
f
v
is
u
al
r
e
p
r
ese
n
t
ati
o
n
s
,
r
at
h
e
r
t
h
a
n
b
ei
n
g
p
ass
i
v
el
y
co
n
ca
t
en
at
ed
.
T
h
e
d
i
r
e
cte
d
at
te
n
ti
o
n
o
p
e
r
a
ti
o
n
i
s
d
ef
in
e
d
as
i
n
(
3
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J
Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
3
,
Sep
tem
b
er
2
0
2
6
:
1
1
1
2
-
1
1
2
2
1116
(
,
,
)
=
(
√
)
(
3
)
W
h
er
e
co
r
r
esp
o
n
d
s
to
lesi
o
n
-
ce
n
tr
ic
v
is
u
al
to
k
en
s
,
an
d
co
r
r
esp
o
n
d
to
h
a
n
d
cr
af
te
d
to
k
en
s
,
an
d
d
en
o
tes
th
e
k
ey
d
im
en
s
io
n
ality
.
R
esid
u
al
co
n
n
ec
tio
n
s
an
d
f
ee
d
-
f
o
r
war
d
r
e
f
in
em
en
t
ar
e
ap
p
lied
to
s
tab
ilize
tr
ain
in
g
an
d
en
h
an
ce
r
ep
r
esen
tatio
n
al
ex
p
r
ess
iv
en
ess
.
T
h
is
f
u
s
io
n
s
tr
ateg
y
en
ab
les
clin
ically
m
ea
n
in
g
f
u
l
cr
o
s
s
-
m
o
d
al
r
ea
s
o
n
in
g
b
y
allo
win
g
h
a
n
d
cr
af
te
d
lesi
o
n
k
n
o
wled
g
e
t
o
d
i
r
ec
tly
in
f
lu
e
n
ce
v
is
u
al
atten
tio
n
d
y
n
am
ics.
Fig
u
r
e
4
.
Dir
ec
ted
tr
an
s
f
o
r
m
er
f
u
s
io
n
m
ec
h
an
is
m
2
.
8
.
Cla
s
s
if
ica
t
io
n a
nd
ca
lib
ra
t
io
n a
s
s
ess
m
ent
T
h
e
f
u
s
ed
r
ep
r
esen
tatio
n
o
b
tain
ed
af
ter
tr
an
s
f
o
r
m
e
r
f
u
s
io
n
is
p
ass
ed
th
r
o
u
g
h
a
d
en
s
e
cla
s
s
if
icatio
n
h
ea
d
f
o
llo
wed
b
y
a
s
o
f
tm
a
x
lay
er
to
p
r
o
d
u
ce
DR
class
p
r
o
b
ab
ilit
ies.
B
ey
o
n
d
s
tan
d
ar
d
d
is
cr
im
in
atio
n
p
er
f
o
r
m
an
ce
,
t
h
e
f
r
am
ew
o
r
k
ex
p
licitly
ev
alu
ates
p
r
e
d
ictio
n
co
n
f
id
en
ce
an
d
ca
lib
r
atio
n
,
as
illu
s
tr
ated
in
Fig
u
r
e
5
.
C
o
n
f
i
d
en
ce
s
co
r
es
a
r
e
an
aly
ze
d
u
s
in
g
ex
p
ec
te
d
ca
lib
r
atio
n
er
r
o
r
(
E
C
E
)
an
d
B
r
ier
s
co
r
e
to
q
u
an
tify
th
e
alig
n
m
en
t
b
etwe
en
p
r
e
d
icted
p
r
o
b
ab
ilit
ies an
d
o
b
s
er
v
ed
ac
cu
r
ac
y
.
Fig
u
r
e
5
.
C
lass
if
icatio
n
an
d
ca
lib
r
atio
n
ev
alu
atio
n
p
ip
elin
e
T
h
is
ca
lib
r
atio
n
-
awa
r
e
e
v
alu
atio
n
en
s
u
r
es
th
at
th
e
p
r
o
p
o
s
ed
s
y
s
tem
n
o
t
o
n
ly
ac
h
i
ev
es
h
ig
h
class
if
icatio
n
p
er
f
o
r
m
an
ce
b
u
t
also
p
r
o
v
id
es
r
eliab
le
c
o
n
f
id
en
ce
esti
m
ates,
wh
ich
ar
e
cr
it
ical
f
o
r
r
ea
l
-
wo
r
ld
DR
s
cr
ee
n
in
g
an
d
r
ef
er
r
al
wo
r
k
f
lo
ws.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
in
teg
r
ates le
s
io
n
ev
id
en
ce
at
b
o
th
th
e
to
k
e
n
g
en
er
atio
n
le
v
el
an
d
th
e
f
u
s
io
n
lev
el,
en
s
u
r
in
g
th
at
p
ath
o
lo
g
ical
in
f
o
r
m
atio
n
ac
tiv
ely
s
h
a
p
es
r
ep
r
esen
tatio
n
lear
n
in
g
.
T
h
e
L
PG
-
T
S
m
ec
h
an
is
m
in
tr
o
d
u
ce
s
lesi
o
n
-
ce
n
tr
ic
to
k
en
izatio
n
,
wh
ile
th
e
d
i
r
ec
ted
tr
an
s
f
o
r
m
er
f
u
s
io
n
en
f
o
r
ce
s
clin
ically
g
u
i
d
ed
cr
o
s
s
-
m
o
d
al
in
te
r
ac
tio
n
.
T
o
g
eth
er
,
th
ese
d
esig
n
ch
o
ice
s
y
ield
a
c
o
h
er
e
n
t,
in
ter
p
r
etab
le,
a
n
d
r
o
b
u
s
t D
R
d
etec
tio
n
f
r
am
ewo
r
k
th
at
alig
n
s
with
p
r
ac
tical
s
cr
ee
n
in
g
r
eq
u
i
r
em
en
ts
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
ex
p
er
im
en
tal
ev
alu
atio
n
o
f
th
e
p
r
o
p
o
s
ed
Han
d
E
f
f
iTr
a
n
s
-
DR
f
r
am
ewo
r
k
is
s
tr
u
ctu
r
ed
to
an
aly
ze
m
o
d
el
b
e
h
av
io
r
at
th
r
ee
c
o
m
p
lem
en
tar
y
lev
els:
i
)
f
o
l
d
-
wis
e
lear
n
in
g
d
y
n
am
ics
an
d
d
is
cr
im
in
ativ
e
p
er
f
o
r
m
an
ce
,
ii)
c
r
o
s
s
-
f
o
ld
r
o
b
u
s
tn
ess
an
d
ca
lib
r
atio
n
co
n
s
is
ten
cy
,
an
d
iii)
g
e
n
er
aliza
tio
n
o
n
i
n
d
ep
e
n
d
en
t
v
alid
atio
n
d
atasets
.
T
h
is
m
u
lti
-
lev
el
an
aly
s
is
en
s
u
r
es
th
at
th
e
r
ep
o
r
ted
p
e
r
f
o
r
m
an
ce
is
n
o
t
c
o
n
f
in
ed
to
a
s
in
g
le
d
ata
s
p
lit
.
I
t in
s
tead
r
ef
lects
s
tab
le
an
d
clin
ically
m
ea
n
in
g
f
u
l
b
eh
av
io
r
ac
r
o
s
s
v
ar
y
in
g
d
ata
d
is
tr
ib
u
tio
n
s
.
3
.
1
.
F
o
ld
-
wis
e
t
ra
ini
ng
dy
na
m
ics a
nd
co
nv
er
g
ence
beha
v
io
r
T
h
e
lear
n
in
g
b
eh
a
v
io
r
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
is
f
ir
s
t
an
aly
ze
d
u
s
in
g
t
h
e
tr
ain
in
g
an
d
v
alid
atio
n
ac
cu
r
ac
y
an
d
lo
s
s
cu
r
v
es
o
b
tai
n
ed
f
r
o
m
f
o
ld
-
1
,
as
s
h
o
wn
in
Fig
u
r
e
6
.
As
o
b
s
er
v
ed
in
Fig
u
r
e
6
(
a)
,
th
e
tr
ain
in
g
an
d
v
alid
atio
n
ac
cu
r
ac
y
c
u
r
v
es
ex
h
ib
it
a
r
a
p
id
i
n
cr
ea
s
e
d
u
r
in
g
t
h
e
in
itial
ep
o
ch
s
,
f
o
llo
wed
b
y
a
g
r
ad
u
al
s
tab
ilizatio
n
,
in
d
i
ca
tin
g
ef
f
ici
en
t
f
ea
tu
r
e
lear
n
in
g
an
d
co
n
v
er
g
en
ce
.
I
m
p
o
r
tan
tly
,
th
e
v
alid
atio
n
ac
cu
r
ac
y
clo
s
ely
f
o
llo
ws
th
e
tr
ain
in
g
ac
cu
r
ac
y
with
o
u
t
s
ig
n
if
ica
n
t
d
iv
er
g
en
ce
,
s
u
g
g
esti
n
g
th
at
th
e
m
o
d
el
av
o
i
d
s
o
v
er
f
itti
n
g
d
esp
ite
its
h
y
b
r
id
a
r
ch
itectu
r
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Les
io
n
-
s
ema
n
tic
to
ke
n
g
u
id
e
d
tr
a
n
s
fo
r
mer fu
s
io
n
fo
r
s
ta
g
e
-
a
d
a
p
tive
d
ia
b
etic
r
etin
o
p
a
th
y
…
(
Mu
r
a
li Gu
jju
la
)
1117
C
o
m
p
lem
en
tar
y
e
v
id
en
ce
is
p
r
o
v
id
e
d
b
y
th
e
lo
s
s
cu
r
v
es
i
n
Fig
u
r
e
6
(
b
)
,
wh
er
e
b
o
t
h
tr
ain
in
g
an
d
v
alid
atio
n
l
o
s
s
es
d
ec
r
ea
s
e
m
o
n
o
to
n
ically
an
d
s
tab
ilize
at
l
o
w
v
alu
es.
T
h
e
a
b
s
en
ce
o
f
o
s
cillato
r
y
b
e
h
av
io
r
o
r
late
-
ep
o
ch
d
iv
e
r
g
en
ce
d
e
m
o
n
s
tr
ates
th
at
th
e
les
io
n
-
p
r
io
r
-
g
u
i
d
ed
to
k
en
izat
io
n
an
d
d
ir
ec
te
d
tr
an
s
f
o
r
m
er
f
u
s
io
n
d
o
n
o
t
in
tr
o
d
u
ce
tr
ain
in
g
i
n
s
tab
ilit
y
,
ev
en
wh
en
m
u
lti
p
le
in
f
o
r
m
atio
n
s
tr
ea
m
s
ar
e
f
u
s
ed
.
T
h
is
s
tab
le
co
n
v
er
g
en
ce
is
p
ar
ticu
la
r
ly
r
elev
an
t
f
o
r
m
ed
ical
im
ag
e
a
n
aly
s
is
,
wh
er
e
o
v
er
-
p
ar
am
ete
r
ized
m
o
d
els
o
f
ten
s
u
f
f
er
f
r
o
m
p
o
o
r
g
en
er
aliza
tio
n
.
(
a)
(
b
)
Fig
u
r
e
6
.
L
ea
r
n
in
g
b
eh
a
v
io
r
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
ac
r
o
s
s
ep
o
ch
s
(
f
o
ld
-
1
)
: (
a)
tr
ai
n
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
y
an
d
(
b
)
tr
ain
in
g
an
d
v
alid
atio
n
lo
s
s
3
.
2
.
F
o
ld
-
wis
e
dis
cr
im
ina
t
iv
e
perf
o
rm
a
nce
a
nd
cla
s
s
-
lev
el
a
na
ly
s
is
T
h
e
class
-
lev
el
d
is
cr
im
in
ativ
e
b
eh
av
i
o
r
o
f
t
h
e
m
o
d
el
f
o
r
f
o
ld
-
1
is
an
aly
ze
d
u
s
in
g
th
e
n
o
r
m
alize
d
co
n
f
u
s
io
n
m
atr
ix
s
h
o
wn
in
Fi
g
u
r
e
7
a
n
d
th
e
c
o
r
r
esp
o
n
d
in
g
o
p
er
atin
g
c
h
ar
ac
ter
is
tics
r
ep
o
r
ted
in
T
ab
le
1
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
r
ev
ea
ls
a
s
tr
o
n
g
d
iag
o
n
al
d
o
m
in
an
ce
,
with
h
ig
h
co
r
r
ec
t
class
if
icatio
n
r
ates
f
o
r
b
o
th
DR
an
d
n
o
-
DR
class
e
s
.
Qu
an
titativ
el
y
,
T
ab
le
1
r
ep
o
r
ts
a
f
o
ld
-
1
a
cc
u
r
ac
y
o
f
0
.
9
5
1
7
,
with
a
s
en
s
itiv
ity
(
r
ec
all)
o
f
0
.
9
7
5
7
f
o
r
t
h
e
DR
class
an
d
a
s
p
ec
if
icity
o
f
0
.
9
1
3
4
f
o
r
th
e
n
o
-
DR
class
.
T
h
is
b
alan
ce
in
d
i
ca
tes
th
at
th
e
m
o
d
el
ef
f
ec
tiv
ely
p
r
io
r
itizes
DR
d
ete
ctio
n
wh
ile
m
ain
tain
in
g
r
ea
s
o
n
ab
le
s
p
ec
if
icity
,
wh
ich
is
ess
en
tial
f
o
r
s
cr
ee
n
in
g
s
ce
n
ar
io
s
wh
er
e
f
alse n
eg
ativ
e
s
ar
e
clin
ically
co
s
tly
.
Fig
u
r
e
7
.
No
r
m
alize
d
co
n
f
u
s
io
n
m
atr
ix
(
f
o
ld
-
1
,
DR
v
s
No
-
D
R
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J
Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
3
,
Sep
tem
b
er
2
0
2
6
:
1
1
1
2
-
1
1
2
2
1118
T
ab
le
1
.
Fo
ld
-
1
o
p
er
atin
g
c
h
ar
ac
ter
is
tics
M
e
t
r
i
c
F
o
r
mu
l
a
(
u
si
n
g
D
R
a
s
p
o
si
t
i
v
e
)
V
a
l
u
e
C
o
u
n
t
s
TP
=
3
6
1
,
F
N
=
9
,
F
P
=
2
0
,
TN
=
2
1
1
,
N
=
601
—
A
c
c
u
r
a
c
y
(
TP
+
TN
)
/
N
0
.
9
5
1
7
S
e
n
s
i
t
i
v
i
t
y
/
R
e
c
a
l
l
(
D
R
)
TP
/
(
TP
+
F
N
)
0
.
9
7
5
7
S
p
e
c
i
f
i
c
i
t
y
(
no
-
D
R
)
TN
/
(
TN
+
FP)
0
.
9
1
3
4
P
r
e
c
i
s
i
o
n
(
D
R
)
TP
/
(
TP
+
FP)
0
.
9
4
7
5
F
1
(
D
R
)
2
·
TP
/
(
2
·
TP
+
FP
+
F
N
)
0
.
9
6
1
4
P
r
e
c
i
s
i
o
n
(
no
-
D
R
)
TN
/
(
TN
+
F
N
)
0
.
9
5
9
1
R
e
c
a
l
l
(
no
-
D
R
)
TN
/
(
TN
+
FP)
0
.
9
1
3
4
F1
-
sc
o
r
e
(
no
-
D
R
)
2
·
TN
/
(
2
·
TN
+
FN
+
FP)
0
.
9
3
5
7
M
a
c
r
o
-
F
1
(
F
o
l
d
-
1)
M
e
a
n
(
F
1
_
D
R
,
F
1
_
N
o
-
D
R
)
0
.
9
4
8
6
T
h
e
ac
h
iev
ed
p
er
f
o
r
m
a
n
ce
o
f
ea
ch
class
alig
n
s
well
with
p
r
ev
io
u
s
lesi
o
n
-
awa
r
e
an
d
atten
t
io
n
-
b
ased
r
esear
ch
in
DR
d
etec
tio
n
,
wh
ich
s
tated
th
at
d
is
cr
im
in
ati
o
n
was
en
h
an
ce
d
th
r
o
u
g
h
h
ig
h
lig
h
tin
g
lesi
o
n
s
[
1
6
]
,
[
2
1
]
.
T
h
is
s
ig
n
if
ies
th
at
th
e
d
ev
elo
p
e
d
f
r
am
ewo
r
k
s
u
cc
ess
f
u
lly
h
ig
h
lig
h
ts
clin
icall
y
im
p
o
r
tan
t
r
etin
al
lesi
o
n
s
.
Fu
r
th
er
in
s
ig
h
t
is
o
b
t
ain
ed
f
r
o
m
t
h
e
class
-
wis
e
F1
s
co
r
es.
T
h
e
DR
class
ac
h
iev
es
an
F1
-
s
co
r
e
o
f
0
.
9
6
1
4
,
wh
ile
th
e
n
o
-
DR
class
attain
s
an
F1
-
s
co
r
e
o
f
0
.
9
3
5
7
,
r
esu
ltin
g
in
a
m
ac
r
o
-
F1
o
f
0
.
9
4
8
6
.
T
h
ese
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
m
o
d
el
d
o
es
n
o
t
ex
h
ib
it
class
b
ias
an
d
m
ain
tain
s
co
n
s
is
ten
t
p
er
f
o
r
m
an
ce
ac
r
o
s
s
b
o
th
p
ath
o
lo
g
ical
an
d
n
o
n
-
p
ath
o
lo
g
ical
s
am
p
les.
T
h
is
b
eh
av
io
r
ca
n
b
e
attr
ib
u
ted
to
th
e
le
s
io
n
-
ce
n
tr
ic
to
k
en
g
en
er
atio
n
,
wh
ich
ex
p
licitly
e
m
p
h
asizes
p
ath
o
lo
g
ical
r
e
g
io
n
s
with
o
u
t su
p
p
r
ess
in
g
g
lo
b
al
r
etin
al
co
n
tex
t.
3
.
3
.
Dis
cr
im
ina
t
iv
e
po
wer
a
nd
pro
ba
bil
i
s
t
ic
re
lia
bil
it
y
B
ey
o
n
d
ac
cu
r
ac
y
-
b
ased
ev
al
u
atio
n
,
th
e
d
is
cr
im
in
ativ
e
ca
p
ab
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
f
r
a
m
ewo
r
k
is
an
aly
ze
d
u
s
in
g
r
ec
ei
v
er
o
p
e
r
at
in
g
ch
ar
ac
ter
is
tic
(
R
OC
)
an
d
p
r
ec
is
io
n
–
r
ec
all
(
PR
)
cu
r
v
es.
T
h
e
R
OC
cu
r
v
e
f
o
r
f
o
ld
-
1
,
as
s
h
o
wn
in
Fig
u
r
e
8
,
y
ield
s
an
AUROC
o
f
ap
p
r
o
x
i
m
ately
0
.
9
9
4
,
in
d
icatin
g
ex
ce
llen
t
class
s
ep
ar
ab
ilit
y
.
Hig
h
AUROC
v
alu
es
ar
e
p
a
r
ticu
lar
ly
im
p
o
r
ta
n
t
in
DR
s
cr
ee
n
i
n
g
,
as
th
ey
r
ef
lect
th
e
m
o
d
el’
s
ab
ilit
y
to
r
an
k
p
ath
o
l
o
g
ical
im
ag
es a
b
o
v
e
n
o
n
-
p
ath
o
l
o
g
ical
o
n
es a
cr
o
s
s
a
wid
e
r
an
g
e
o
f
d
ec
is
io
n
th
r
esh
o
ld
s
.
Fig
u
r
e
8
.
R
OC
cu
r
v
e
with
AU
R
OC
(
f
o
ld
-
1)
E
v
alu
atio
n
r
esu
lts
f
o
r
th
e
Han
d
E
f
f
iTr
a
n
s
-
DR
alg
o
r
ith
m
in
ter
m
s
o
f
d
is
cr
im
in
atio
n
q
u
ality
an
d
p
r
o
b
a
b
ilit
y
ca
lib
r
atio
n
ar
e
s
h
o
wn
in
Fig
u
r
e
9
.
I
n
Fig
u
r
e
9
(
a
)
,
th
e
PR
cu
r
v
e
is
d
ep
icted
,
w
h
er
ea
s
Fig
u
r
e
9
(
b
)
d
is
p
lay
s
th
e
r
eliab
ilit
y
d
iag
r
am
.
T
h
e
ar
ea
u
n
d
e
r
th
e
p
r
ec
is
io
n
–
r
ec
all
cu
r
v
e
(
AUPR
C
)
is
q
u
ite
h
ig
h
(
AUPR
C
≈
0
.
9
9
4
)
.
T
h
e
latter
f
ac
t
in
d
icate
s
g
o
o
d
tr
ad
e
-
o
f
f
PR
p
r
o
p
er
ties
f
o
r
th
e
DR
clas
s
.
T
h
e
m
o
r
e
im
p
o
r
tan
t
p
o
in
t
is
th
at
t
h
e
r
eliab
ilit
y
d
iag
r
am
in
d
ica
tes
a
g
o
o
d
m
atch
b
etwe
en
p
r
ed
ic
ted
p
r
o
b
ab
ilit
y
an
d
em
p
ir
ical
ac
cu
r
ac
y
.
T
h
is
o
b
s
e
r
v
atio
n
is
q
u
an
titativ
ely
s
u
p
p
o
r
ted
b
y
th
e
d
is
cr
im
in
ativ
e
an
d
ca
lib
r
atio
n
m
etr
ics
r
ep
o
r
ted
in
T
ab
le
2
,
wh
er
e
th
e
E
C
E
an
d
B
r
ier
s
co
r
e
r
em
ain
lo
w
(
0
.
0
2
5
an
d
0
.
0
5
8
,
r
esp
ec
tiv
ely
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Les
io
n
-
s
ema
n
tic
to
ke
n
g
u
id
e
d
tr
a
n
s
fo
r
mer fu
s
io
n
fo
r
s
ta
g
e
-
a
d
a
p
tive
d
ia
b
etic
r
etin
o
p
a
th
y
…
(
Mu
r
a
li Gu
jju
la
)
1119
Su
ch
ca
lib
r
atio
n
b
eh
av
i
o
r
is
cr
itical
f
o
r
clin
ical
d
ec
is
io
n
s
u
p
p
o
r
t
s
y
s
tem
s
,
wh
er
e
p
r
ed
icted
co
n
f
id
en
ce
o
f
ten
in
f
o
r
m
s
r
ef
er
r
al
th
r
esh
o
ld
s
an
d
f
o
llo
w
-
u
p
ac
tio
n
s
.
T
h
e
AU
R
OC
an
d
ca
lib
r
atio
n
m
etr
ics
ar
e
s
im
ilar
to
o
th
er
s
tate
-
of
-
th
e
-
ar
t
DR
d
etec
to
r
b
ased
o
n
tr
an
s
f
o
r
m
e
r
s
[
1
7
]
,
a
n
d
th
e
E
C
E
an
d
B
r
ier
s
co
r
es
ar
e
co
n
s
is
ten
t
with
r
ec
e
n
t
s
tu
d
ies
em
p
h
asizin
g
th
e
n
ec
ess
ity
o
f
a
well
-
ca
lib
r
ated
d
ee
p
lear
n
in
g
m
o
d
el
in
h
ea
lth
ca
r
e
d
ec
is
io
n
-
m
ak
in
g
s
y
s
tem
s
[
2
0
]
,
[
2
5
]
.
(
a)
(
b
)
Fig
u
r
e
9
.
Per
f
o
r
m
an
c
e
ev
alu
at
io
n
o
f
t
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
: (
a)
PR
cu
r
v
e
a
n
d
(
b
)
r
eliab
il
ity
(
ca
lib
r
atio
n
)
T
ab
le
2
.
Fo
ld
-
1
d
is
cr
im
in
ativ
e
an
d
ca
lib
r
atio
n
m
etr
ics
M
e
t
r
i
c
V
a
l
u
e
A
U
R
O
C
0
.
9
9
4
A
U
P
R
C
0
.
9
9
4
EC
E
0
.
0
2
5
B
r
i
e
r
s
c
o
r
e
0
.
0
5
8
3
.
4
.
Cro
s
s
-
f
o
ld ro
bu
s
t
nes
s
a
nd
s
t
a
bil
it
y
a
na
ly
s
is
T
o
ass
ess
r
o
b
u
s
tn
ess
ac
r
o
s
s
d
if
f
er
en
t
d
ata
p
ar
titi
o
n
s
,
f
iv
e
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
r
esu
lts
ar
e
s
u
m
m
ar
ized
in
T
ab
le
3
.
T
h
e
m
ea
n
ac
cu
r
ac
y
ac
r
o
s
s
f
iv
e
f
o
ld
s
is
0
.
9
6
5
3
with
a
s
tan
d
ar
d
d
ev
iatio
n
o
f
0
.
0
0
8
3
,
wh
ile
th
e
m
ea
n
m
ac
r
o
-
F1
s
c
o
r
e
is
0
.
9
6
5
3
with
a
s
im
ilar
ly
lo
w
v
ar
ian
ce
.
T
h
ese
r
esu
lt
s
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
f
r
a
m
ewo
r
k
m
ain
tai
n
s
s
tab
le
p
er
f
o
r
m
an
ce
ac
r
o
s
s
f
o
ld
s
an
d
is
n
o
t
o
v
er
ly
s
en
s
itiv
e
to
s
p
ec
if
ic
tr
ain
–
test
s
p
lits
.
C
o
n
s
is
ten
tly
h
ig
h
AURO
C
(
m
ea
n
0
.
9
9
3
1
)
a
n
d
AUPR
C
(
m
ea
n
0
.
9
9
3
7
)
v
alu
es
ac
r
o
s
s
f
o
ld
s
f
u
r
th
er
co
n
f
ir
m
t
h
e
r
o
b
u
s
tn
ess
o
f
th
e
d
is
cr
im
in
ativ
e
ca
p
a
b
ilit
y
.
T
h
e
r
elativ
ely
lo
w
s
tan
d
ar
d
d
e
v
iatio
n
o
f
C
o
h
en
’
s
k
ap
p
a
(
0
.
0
1
6
7
)
s
u
g
g
ests
s
tr
o
n
g
ag
r
ee
m
e
n
t
b
ey
o
n
d
ch
an
ce
a
cr
o
s
s
all
f
o
ld
s
,
r
ein
f
o
r
cin
g
t
h
e
r
eliab
ilit
y
o
f
th
e
m
o
d
el.
I
m
p
o
r
tan
tly
,
ca
lib
r
atio
n
m
etr
ics
s
u
c
h
as
E
C
E
an
d
B
r
ier
s
co
r
e
r
em
ain
l
o
w
an
d
s
tab
le
ac
r
o
s
s
f
o
ld
s
,
d
em
o
n
s
tr
atin
g
t
h
at
p
r
o
b
ab
ilit
y
esti
m
ates a
r
e
co
n
s
is
ten
tly
r
eliab
le
an
d
n
o
t c
o
n
f
in
ed
t
o
a
s
in
g
le
f
av
o
r
a
b
l
e
s
p
lit.
T
ab
le
3
.
C
r
o
s
s
-
f
o
ld
p
er
f
o
r
m
an
ce
s
u
m
m
ar
y
(
5
f
o
ld
s
)
M
e
t
r
i
c
M
e
a
n
S
t
d
.
d
e
v
i
a
t
i
o
n
A
c
c
u
r
a
c
y
0
.
9
6
5
3
1
8
0
.
0
0
8
3
4
5
M
a
c
r
o
-
F1
0
.
9
6
5
3
0
3
0
.
0
0
8
3
4
6
A
U
R
O
C
0
.
9
9
3
0
9
4
0
.
0
0
2
0
9
A
U
P
R
C
0
.
9
9
3
7
1
7
0
.
0
0
1
3
8
2
C
o
h
e
n
’
s
k
a
p
p
a
0
.
9
3
0
6
2
0
.
0
1
6
6
7
9
EC
E
0
.
0
2
4
9
4
5
0
.
0
0
7
4
9
4
B
r
i
e
r
s
c
o
r
e
0
.
0
5
7
8
4
0
.
0
1
2
6
2
8
3
.
5
.
E
x
t
er
na
l
v
a
lid
a
t
io
n a
nd
g
ener
a
liza
t
io
n
ca
pa
bil
it
y
T
o
ev
alu
ate
g
en
e
r
aliza
tio
n
b
ey
o
n
d
th
e
p
r
im
a
r
y
d
ataset,
th
e
tr
ain
e
d
m
o
d
el
is
test
e
d
o
n
two
in
d
ep
en
d
en
t v
alid
atio
n
d
atasets
.
T
h
e
p
er
f
o
r
m
an
ce
o
n
v
alid
ati
o
n
_
d
ataset
-
1
is
r
ep
o
r
ted
in
T
a
b
le
4
,
wh
ile
r
esu
lts
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J
Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
3
,
Sep
tem
b
er
2
0
2
6
:
1
1
1
2
-
1
1
2
2
1120
on
v
alid
atio
n
_
d
ataset
-
2
ar
e
p
r
esen
ted
in
T
ab
le
5
.
On
v
alid
atio
n
_
d
ataset
-
1
,
th
e
m
o
d
el
ac
h
iev
es
a
m
ea
n
ac
cu
r
ac
y
o
f
0
.
9
5
6
8
an
d
a
m
ac
r
o
-
F1
-
s
co
r
e
o
f
0
.
9
5
6
0
,
wit
h
AURO
C
an
d
AUP
R
C
v
alu
es
ex
ce
ed
in
g
0
.
9
9
.
Alth
o
u
g
h
a
s
lig
h
t
r
ed
u
ctio
n
i
s
o
b
s
er
v
ed
co
m
p
ar
ed
to
t
h
e
p
r
im
ar
y
d
ataset,
th
e
p
er
f
o
r
m
an
ce
r
em
ain
s
s
tr
o
n
g
,
in
d
icatin
g
ef
f
ec
tiv
e
tr
an
s
f
er
ab
i
lity
.
Similar
ly
,
v
alid
atio
n
_
d
ataset
-
2
y
ield
s
a
m
ea
n
ac
cu
r
ac
y
o
f
0
.
9
6
2
7
an
d
m
ac
r
o
-
F1
o
f
0
.
9
6
2
1
,
with
AURO
C
an
d
AUPR
C
v
alu
es
o
f
0
.
9
9
2
4
a
n
d
0
.
9
9
2
8
,
r
esp
ec
tiv
ely
.
T
h
e
ca
lib
r
atio
n
m
etr
ics
o
n
b
o
th
v
alid
atio
n
d
atasets
r
em
ain
c
o
m
p
ar
ab
le
to
t
h
o
s
e
o
b
s
er
v
ed
d
u
r
in
g
cr
o
s
s
-
v
alid
atio
n
,
s
u
g
g
esti
n
g
th
at
th
e
m
o
d
el
d
o
es
n
o
t
e
x
h
ib
it
o
v
er
co
n
f
id
e
n
t
p
r
ed
ictio
n
s
wh
en
ex
p
o
s
ed
to
u
n
s
ee
n
d
ata
d
is
tr
ib
u
tio
n
s
.
T
h
is
b
eh
a
v
io
r
is
p
ar
ticu
lar
ly
im
p
o
r
tan
t
f
o
r
r
e
al
-
wo
r
ld
d
e
p
lo
y
m
e
n
t,
wh
er
e
v
ar
iatio
n
s
in
ac
q
u
is
itio
n
d
ev
ices
an
d
p
atien
t
d
em
o
g
r
a
p
h
ics
ar
e
in
e
v
itab
le.
T
ab
le
4
.
Valid
atio
n
p
er
f
o
r
m
an
ce
o
n
v
alid
atio
n
_
d
ataset
-
1
M
e
t
r
i
c
M
e
a
n
S
t
d
.
d
e
v
i
a
t
i
o
n
A
c
c
u
r
a
c
y
0
.
9
5
6
8
4
0
.
0
0
9
2
1
5
M
a
c
r
o
-
F1
0
.
9
5
5
9
9
0
.
0
0
9
4
8
7
A
U
R
O
C
0
.
9
9
0
7
5
0
.
0
0
2
6
4
1
A
U
P
R
C
0
.
9
9
1
1
0
.
0
0
2
0
7
8
C
o
h
e
n
’
s
k
a
p
p
a
0
.
9
1
5
6
7
0
.
0
1
8
3
9
2
EC
E
0
.
0
2
8
9
1
0
.
0
0
6
9
2
8
B
r
i
e
r
S
c
o
r
e
0
.
0
6
2
3
4
0
.
0
1
1
8
5
3
T
ab
le
5
.
Valid
atio
n
p
er
f
o
r
m
an
ce
o
n
v
alid
atio
n
_
d
ataset
-
2
M
e
t
r
i
c
M
e
a
n
S
t
d
.
d
e
v
i
a
t
i
o
n
A
c
c
u
r
a
c
y
0
.
9
6
2
7
3
0
.
0
0
8
7
4
9
M
a
c
r
o
-
F1
0
.
9
6
2
1
4
0
.
0
0
8
6
8
1
A
U
R
O
C
0
.
9
9
2
4
1
0
.
0
0
2
3
7
1
A
U
P
R
C
0
.
9
9
2
8
2
0
.
0
0
1
7
3
1
C
o
h
e
n
’
s
k
a
p
p
a
0
.
9
2
4
5
8
0
.
0
1
7
1
5
6
EC
E
0
.
0
2
6
1
2
0
.
0
0
6
8
4
5
B
r
i
e
r
S
c
o
r
e
0
.
0
5
9
1
0
.
0
1
2
0
0
1
3
.
6
.
Co
m
pa
ra
t
iv
e
perf
o
r
m
a
nce
a
cr
o
s
s
da
t
a
s
et
s
A
co
n
s
o
lid
ated
co
m
p
ar
is
o
n
o
f
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
th
e
p
r
im
a
r
y
d
ataset
an
d
b
o
th
v
ali
d
atio
n
d
atasets
i
s
p
r
o
v
id
e
d
i
n
T
a
b
le
6
.
T
h
e
t
ab
le
h
ig
h
lig
h
ts
th
at
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
m
ain
tai
n
s
co
n
s
is
ten
tly
h
i
g
h
p
er
f
o
r
m
an
ce
ac
r
o
s
s
all
d
atasets
,
with
ac
cu
r
ac
y
an
d
m
ac
r
o
-
F
1
v
alu
es
r
em
ain
i
n
g
ab
o
v
e
0
.
9
5
an
d
d
is
cr
im
in
ativ
e
m
etr
ics
co
n
s
is
ten
tly
ex
ce
ed
i
n
g
0
.
9
9
.
W
h
ile
a
m
o
d
est
d
eg
r
a
d
atio
n
is
o
b
s
er
v
ed
o
n
e
x
ter
n
al
d
atasets
,
th
e
d
r
o
p
is
lim
ited
an
d
ex
p
ec
ted
,
in
d
icatin
g
s
tr
o
n
g
g
en
e
r
aliza
tio
n
r
ath
er
th
an
d
ataset
-
s
p
ec
if
ic
o
p
tim
iza
tio
n
.
T
ab
le
6
.
Ov
e
r
all
av
er
ag
e
d
p
er
f
o
r
m
an
ce
c
o
m
p
ar
is
o
n
ac
r
o
s
s
d
atasets
D
a
t
a
s
e
t
A
c
c
u
r
a
c
y
M
a
c
r
o
-
F1
A
U
R
O
C
A
U
P
R
C
C
o
h
e
n
’
s
k
a
p
p
a
EC
E
B
r
i
e
r
s
c
o
r
e
P
r
i
mary
d
a
t
a
set
(
5
-
f
o
l
d
)
0
.
9
6
5
3
0
.
9
6
5
3
0
.
9
9
3
1
0
.
9
9
3
7
0
.
9
3
0
6
0
.
0
2
4
9
0
.
0
5
7
8
V
a
l
i
d
a
t
i
o
n
d
a
t
a
se
t
-
1
0
.
9
5
6
8
0
.
9
5
6
0
.
9
9
0
8
0
.
9
9
1
1
0
.
9
1
5
7
0
.
0
2
8
9
0
.
0
6
2
3
V
a
l
i
d
a
t
i
o
n
d
a
t
a
se
t
-
2
0
.
9
6
2
7
0
.
9
6
2
1
0
.
9
9
2
4
0
.
9
9
2
8
0
.
9
2
4
6
0
.
0
2
6
1
0
.
0
5
9
1
Ov
er
all,
th
ese
f
in
d
in
g
s
ar
e
co
n
s
is
ten
t
with
r
ec
en
t
C
N
N
-
tr
an
s
f
o
r
m
er
h
y
b
r
id
f
r
am
ewo
r
k
s
f
o
r
r
etin
al
d
is
ea
s
e
an
aly
s
is
[
1
4
]
,
[
1
7
]
.
T
h
e
p
r
o
p
o
s
ed
Ha
n
d
E
f
f
iT
r
an
s
-
D
R
f
r
am
ewo
r
k
f
u
r
th
er
s
tr
en
g
th
e
n
s
th
ese
ap
p
r
o
ac
h
es
b
y
in
teg
r
atin
g
lesi
o
n
-
g
u
id
ed
t
o
k
en
izatio
n
with
d
ir
ec
ted
tr
a
n
s
f
o
r
m
er
f
u
s
io
n
,
r
esu
ltin
g
in
r
o
b
u
s
t
an
d
r
eliab
le
DR
d
etec
tio
n
.
T
h
e
f
o
ld
-
wis
e
s
tab
ilit
y
,
co
m
b
in
ed
with
s
tr
o
n
g
ex
ter
n
al
v
alid
atio
n
p
er
f
o
r
m
an
ce
,
s
u
p
p
o
r
ts
th
e
f
r
am
ewo
r
k
’
s
s
u
itab
ilit
y
f
o
r
DR
s
cr
ee
n
in
g
s
ce
n
ar
io
s
wh
er
e
co
n
s
is
ten
t
b
eh
av
io
r
ac
r
o
s
s
p
o
p
u
latio
n
s
an
d
r
eliab
le
co
n
f
id
en
ce
esti
m
atio
n
ar
e
ess
en
tial.
4.
CO
NCLU
SI
O
N
T
h
is
wo
r
k
p
r
esen
ted
Han
d
E
f
f
iTr
an
s
-
DR
,
a
lesi
o
n
-
awa
r
e
h
y
b
r
i
d
f
r
a
m
ewo
r
k
f
o
r
au
to
m
ated
DR
d
etec
tio
n
th
at
e
x
p
licitly
in
teg
r
ates
p
ath
o
lo
g
ical
p
r
io
r
s
i
n
to
b
o
th
r
ep
r
esen
tatio
n
lear
n
i
n
g
an
d
f
ea
tu
r
e
f
u
s
io
n
s
tag
es.
B
y
in
tr
o
d
u
cin
g
a
le
s
io
n
-
g
u
id
ed
to
k
en
izatio
n
s
tr
ateg
y
an
d
a
d
ir
ec
ted
tr
a
n
s
f
o
r
m
er
-
b
ased
f
u
s
io
n
m
ec
h
an
is
m
,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
e
n
s
u
r
es
th
a
t
clin
ically
r
elev
an
t
lesi
o
n
in
f
o
r
m
atio
n
ac
t
iv
ely
in
f
lu
e
n
ce
s
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
Les
io
n
-
s
ema
n
tic
to
ke
n
g
u
id
e
d
tr
a
n
s
fo
r
mer fu
s
io
n
fo
r
s
ta
g
e
-
a
d
a
p
tive
d
ia
b
etic
r
etin
o
p
a
th
y
…
(
Mu
r
a
li Gu
jju
la
)
1121
lear
n
ed
v
is
u
al
r
ep
r
esen
tatio
n
s
.
T
h
e
u
s
e
o
f
s
tr
atif
ied
c
r
o
s
s
-
v
alid
atio
n
an
d
e
x
ter
n
al
d
ataset
v
alid
atio
n
d
em
o
n
s
tr
ates
r
o
b
u
s
tn
ess
u
n
d
e
r
class
im
b
alan
ce
an
d
d
ataset
v
ar
iab
ilit
y
,
wh
ile
ca
lib
r
atio
n
an
aly
s
is
h
ig
h
lig
h
ts
th
e
r
eliab
ilit
y
o
f
p
r
ed
ictio
n
co
n
f
id
en
ce
.
Ov
er
all,
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
ad
v
an
ce
s
au
to
m
ated
DR
s
cr
ee
n
in
g
b
y
co
m
b
in
in
g
ac
c
u
r
ac
y
,
r
o
b
u
s
tn
ess
,
an
d
in
ter
p
r
eta
b
ilit
y
,
m
ak
in
g
it
s
u
itab
le
f
o
r
lar
g
e
-
s
ca
le
clin
ical
d
ep
lo
y
m
e
n
t a
n
d
d
ec
is
io
n
-
s
u
p
p
o
r
t sy
s
tem
s
.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
wo
u
ld
lik
e
to
ac
k
n
o
wled
g
e
Vels
I
n
s
titu
te
o
f
Scien
ce
,
T
ec
h
n
o
lo
g
y
an
d
Ad
v
an
ce
d
Stu
d
ies,
C
h
en
n
ai,
I
n
d
ia,
f
o
r
p
r
o
v
id
in
g
th
e
n
ec
ess
ar
y
ac
ad
e
m
ic
an
d
r
esear
ch
s
u
p
p
o
r
t f
o
r
th
is
s
tu
d
y
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
r
ec
eiv
ed
n
o
s
p
ec
if
ic
g
r
an
t
f
r
o
m
a
n
y
f
u
n
d
in
g
ag
en
c
y
in
th
e
p
u
b
lic,
c
o
m
m
er
cial,
o
r
not
-
f
o
r
-
p
r
o
f
it secto
r
s
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Mu
r
ali
Gu
jju
la
✓
✓
✓
✓
✓
✓
✓
✓
Ku
m
ar
Nar
ay
an
an
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
d
ec
lar
e
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
p
r
i
m
ar
y
r
etin
al
f
u
n
d
u
s
i
m
ag
es
u
s
ed
in
th
is
s
tu
d
y
wer
e
o
b
tain
e
d
f
r
o
m
th
e
p
u
b
licly
ac
ce
s
s
ib
l
e
d
ataset
o
n
Ka
g
g
le
at
h
ttp
s
://www.
k
ag
g
le.
co
m
/c/ap
to
s
2
0
1
9
-
b
lin
d
n
ess
-
d
etec
tio
n
/d
ata.
T
h
e
s
o
u
r
ce
tr
ain
in
g
s
et
co
n
tain
s
3
,
6
6
2
la
b
elled
im
ag
e
s
:
1
,
8
0
5
no
_
DR
im
ag
es
an
d
1
,
8
5
7
DR
im
ag
es
af
ter
co
m
b
in
in
g
th
e
m
ild
(
3
7
0
)
,
m
o
d
er
ate
(
9
9
9
)
,
s
ev
er
e
(
1
9
3
)
,
an
d
p
r
o
life
r
ativ
e
DR
(
2
9
5
)
c
ateg
o
r
ies.
No
n
ew
p
atien
t
o
r
clin
ical
d
ata
wer
e
co
llected
.
Acc
ess
to
th
e
s
o
u
r
ce
im
ag
es
r
eq
u
ir
es
co
m
p
li
an
ce
with
Kag
g
le
r
eg
is
tr
atio
n
an
d
th
e
d
ataset
p
r
o
v
id
e
r
’
s
ter
m
s
; th
er
ef
o
r
e
,
th
e
im
ag
es a
r
e
n
o
t r
ed
is
tr
ib
u
ted
with
th
is
ar
ticle.
Ag
g
r
eg
ate
r
esu
lts
ar
e
r
ep
o
r
ted
in
th
e
ar
ticle.
Stu
d
y
-
g
en
er
ated
s
p
lit
f
iles
an
d
p
r
o
ce
s
s
ed
o
u
tp
u
ts
ar
e
av
ailab
le
f
r
o
m
th
e
c
o
r
r
esp
o
n
d
i
n
g
au
t
h
o
r
,
[
MG
]
,
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
est,
s
u
b
ject
to
th
e
o
r
ig
in
al
d
atase
t te
r
m
s
RE
F
E
R
E
NC
E
S
[
1
]
T.
Y
.
W
o
n
g
,
C
.
M
.
G
.
C
h
e
u
n
g
,
M
.
La
r
se
n
,
S
.
S
h
a
r
ma,
a
n
d
R
.
S
i
mó
,
“
D
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
,
”
N
a
t
u
re
R
e
v
i
e
w
s
D
i
sea
se
Pri
m
e
rs
,
v
o
l
.
2
,
n
o
.
1
,
M
a
r
.
2
0
1
6
,
d
o
i
:
1
0
.
1
0
3
8
/
n
r
d
p
.
2
0
1
6
.
1
2
.
[
2
]
M
.
M
.
I
sl
a
m,
H
.
-
C
.
Y
a
n
g
,
T.
N
.
P
o
l
y
,
W
.
-
S
.
J
i
a
n
,
a
n
d
Y
.
-
C
.
J
.
L
i
,
“
D
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
f
o
r
d
e
t
e
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
i
n
r
e
t
i
n
a
l
f
u
n
d
u
s
p
h
o
t
o
g
r
a
p
h
s:
a
s
y
st
e
ma
t
i
c
r
e
v
i
e
w
a
n
d
m
e
t
a
-
a
n
a
l
y
s
i
s,”
C
o
m
p
u
t
e
r
Me
t
h
o
d
s
a
n
d
Pro
g
r
a
m
s
i
n
Bi
o
m
e
d
i
c
i
n
e
,
v
o
l
.
1
9
1
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
m
p
b
.
2
0
2
0
.
1
0
5
3
2
0
.
[
3
]
V
.
G
u
l
s
h
a
n
e
t
a
l
.
,
“
D
e
v
e
l
o
p
m
e
n
t
a
n
d
v
a
l
i
d
a
t
i
o
n
o
f
a
d
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
d
e
t
e
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
i
n
r
e
t
i
n
a
l
f
u
n
d
u
s
p
h
o
t
o
g
r
a
p
h
s
,
”
J
A
MA
,
v
o
l
.
3
1
6
,
n
o
.
2
2
,
D
e
c
.
2
0
1
6
,
d
o
i
:
1
0
.
1
0
0
1
/
j
a
ma.
2
0
1
6
.
1
7
2
1
6
.
[
4
]
D
.
S
.
W
.
T
i
n
g
e
t
a
l
.
,
“
D
e
v
e
l
o
p
m
e
n
t
a
n
d
v
a
l
i
d
a
t
i
o
n
o
f
a
d
e
e
p
l
e
a
r
n
i
n
g
sy
s
t
e
m
f
o
r
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
a
n
d
r
e
l
a
t
e
d
e
y
e
d
i
s
e
a
se
s
u
si
n
g
r
e
t
i
n
a
l
i
m
a
g
e
s
f
r
o
m
m
u
l
t
i
e
t
h
n
i
c
p
o
p
u
l
a
t
i
o
n
s
w
i
t
h
d
i
a
b
e
t
e
s,
”
J
A
MA
,
v
o
l
.
3
1
8
,
n
o
.
2
2
,
D
e
c
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
0
1
/
j
a
m
a
.
2
0
1
7
.
1
8
1
5
2
.
[
5
]
M
.
D
.
A
b
r
à
mo
f
f
,
P
.
T.
La
v
i
n
,
M
.
B
i
r
c
h
,
N
.
S
h
a
h
,
a
n
d
J
.
C
.
F
o
l
k
,
“
P
i
v
o
t
a
l
t
r
i
a
l
o
f
a
n
a
u
t
o
n
o
mo
u
s
A
I
-
b
a
se
d
d
i
a
g
n
o
s
t
i
c
s
y
st
e
m
f
o
r
d
e
t
e
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
i
n
p
r
i
mar
y
c
a
r
e
o
f
f
i
c
e
s,
”
n
p
j
D
i
g
i
t
a
l
M
e
d
i
c
i
n
e
,
v
o
l
.
1
,
n
o
.
1
,
A
u
g
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
7
4
6
-
018
-
0
0
4
0
-
6.
[6
]
M
.
V
o
e
t
s,
K
.
M
ø
l
l
e
r
se
n
,
a
n
d
L.
A
.
B
o
n
g
o
,
“
R
e
p
r
o
d
u
c
t
i
o
n
s
t
u
d
y
u
s
i
n
g
p
u
b
l
i
c
d
a
t
a
o
f
:
d
e
v
e
l
o
p
m
e
n
t
a
n
d
v
a
l
i
d
a
t
i
o
n
o
f
a
d
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
d
e
t
e
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
i
n
r
e
t
i
n
a
l
f
u
n
d
u
s
p
h
o
t
o
g
r
a
p
h
s
,
”
PLO
S
O
N
E
,
v
o
l
.
1
4
,
n
o
.
6
,
J
u
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
3
7
1
/
j
o
u
r
n
a
l
.
p
o
n
e
.
0
2
1
7
5
4
1
.
[
7
]
N
.
Tsi
k
n
a
k
i
s
e
t
a
l
.
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
d
i
a
b
e
t
i
c
r
e
t
i
n
o
p
a
t
h
y
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
a
s
e
d
o
n
f
u
n
d
u
s
i
m
a
g
e
s:
a
r
e
v
i
e
w
,
”
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
.
1
3
5
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
b
i
o
me
d
.
2
0
2
1
.
1
0
4
5
9
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.