I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
14
,
No
.
5
,
Octo
b
er
2
0
2
5
,
p
p
.
4
0
3
2
~
4
0
4
2
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
5
.
p
p
4
0
3
2
-
4
0
4
2
4032
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Residua
l edge d
e
nse enha
nced
mo
dule ne
twork
:
a
d
eep learning
a
ppro
a
ch wit
h
m
ulti
-
cla
ss
S
VM
f
o
r lung tu
mo
r stag
e
cla
ss
ificatio
n
P
ra
ba
k
a
r
a
n J
a
y
a
ra
m
a
n
1
,
P
a
nd
ia
ra
j
Selv
a
ra
j
2
,
Ash
wini E
la
ng
o
3
1
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,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
a
n
d
Te
c
h
n
o
l
o
g
y
,
S
R
M
I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
a
n
d
T
e
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
C
o
m
p
u
t
i
n
g
T
e
c
h
n
o
l
o
g
i
e
s,
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
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
A
n
a
e
st
h
e
si
a
,
V
e
l
a
m
m
a
l
M
e
d
i
c
a
l
C
o
l
l
e
g
e
H
o
sp
i
t
a
l
a
n
d
R
e
se
a
r
c
h
I
n
st
i
t
u
t
e
,
M
a
d
u
r
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
J
u
n
6
,
2024
R
ev
is
ed
J
u
n
2
5
,
2
0
2
5
Acc
ep
ted
J
u
l 1
3
,
2
0
2
5
Lu
n
g
c
a
n
c
e
r
se
g
m
e
n
tatio
n
wi
th
p
o
sitr
o
n
e
m
issio
n
to
m
o
g
ra
p
h
y
(
P
ET
)
a
n
d
c
o
m
p
u
ted
to
m
o
g
ra
p
h
y
(CT)
ima
g
e
s
p
lay
s
a
c
rit
ica
l
ro
le
to
a
c
c
u
ra
tely
d
e
tec
t
lu
n
g
c
a
n
c
e
r.
Ne
v
e
rt
h
e
les
s,
lu
n
g
t
u
m
o
r
se
g
m
e
n
tati
o
n
i
n
P
ET
/CT
i
m
a
g
e
s
we
re
e
x
trem
e
ly
d
iff
icu
lt
d
u
e
t
o
t
h
e
m
o
v
e
m
e
n
t
c
a
u
se
d
b
y
re
sp
irati
o
n
.
D
e
sp
it
e
th
is
fa
c
t,
th
e
lu
n
g
t
u
m
o
r
ima
g
e
s
sh
o
w
n
larg
e
n
u
m
b
e
r
o
f
v
a
riati
o
n
s
m
o
st
ly
in
P
ET
ima
g
e
s
a
n
d
CT
ima
g
e
s.
As
P
ET
-
CT
ima
g
e
s
a
re
a
c
q
u
ired
c
o
n
c
u
r
re
n
tl
y
t
h
e
sh
a
p
e
a
n
d
siz
e
o
f
lu
n
g
t
u
m
o
r
v
a
ri
e
s
a
c
c
o
rd
in
g
to
m
o
d
a
li
t
y
.
To
a
d
d
re
ss
th
e
se
issu
e
s,
we
d
e
v
e
lo
p
e
d
a
re
sid
u
a
l
e
d
g
e
d
e
n
se
e
n
h
a
n
c
e
d
m
o
d
u
le
n
e
two
rk
(
REDEM
-
NET
)
fra
m
e
wo
rk
fo
r
lu
n
g
tu
m
o
r
sta
g
e
c
las
sifica
ti
o
n
.
Th
e
p
ro
p
o
se
d
REDEM
-
NET
c
a
n
p
r
o
c
e
ss
P
ET
a
n
d
CT
ima
g
e
s
a
s
in
p
u
ts.
In
a
d
d
it
i
o
n
,
t
h
e
d
e
n
se
re
sid
u
a
l
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
two
r
k
(DRCN
)
c
o
ll
e
c
ts
b
o
th
in
p
u
ts
a
n
d
e
x
trac
ts
h
ig
h
-
d
ime
n
s
io
n
a
l
fe
a
tu
re
s
c
o
n
c
u
rre
n
tl
y
.
T
h
e
e
x
trac
ted
fe
a
tu
re
s
fro
m
b
o
t
h
ima
g
i
n
g
m
o
d
a
li
ti
e
s
we
re
fe
d
in
t
o
UN
e
t++
+
to
o
b
tain
m
u
lt
i
-
lev
e
l
d
e
c
o
d
e
d
fe
a
tu
re
s.
Th
e
e
x
trac
ted
d
e
c
o
d
e
d
fe
a
tu
re
s
a
re
c
o
n
c
u
rre
n
tl
y
s
u
p
p
li
e
d
to
t
h
e
p
ix
e
l
lev
e
l
lea
rn
i
n
g
m
o
d
u
le
(P
EL
M
)
a
n
d
e
d
g
e
lev
e
l
lea
rn
in
g
m
o
d
u
le
(E2
LM
)
w
h
ich
re
su
l
ti
n
g
i
n
tw
o
o
u
t
p
u
ts
fo
r
s
u
b
se
q
u
e
n
t
lea
rn
in
g
.
Th
e
o
u
tp
u
ts
we
re
m
e
r
g
e
d
to
p
r
o
v
id
e
a
v
e
ry
p
re
c
ise
lu
n
g
t
u
m
o
r
se
g
m
e
n
tatio
n
.
F
u
rth
e
rm
o
re
,
se
g
m
e
n
ted
t
u
m
o
r
wa
s
fe
d
t
o
m
u
lt
i
-
c
la
ss
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
(M
C
-
S
V
M
)
fo
r
l
u
n
g
tu
m
o
r
sta
g
e
c
las
sifica
ti
o
n
.
M
o
re
o
v
e
r,
it
wa
s
a
b
le
to
id
e
n
ti
fy
th
re
e
sta
g
e
s
a
n
d
it
s
s
u
b
sta
g
e
s
n
a
m
e
ly
p
rima
ry
tu
m
o
r,
re
g
io
n
ly
m
p
h
n
o
d
e
a
n
d
d
istan
t
m
e
tas
tas
is.
K
ey
w
o
r
d
s
:
C
o
m
p
u
ted
to
m
o
g
r
a
p
h
y
Dee
p
lear
n
in
g
L
u
n
g
ca
n
ce
r
Ma
ch
in
e
lear
n
in
g
M
u
l
t
i
-
c
l
a
s
s
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
Po
s
itro
n
em
is
s
io
n
tech
n
o
lo
g
y
R
E
DE
M
-
NE
T
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
:
Pan
d
iar
aj
Selv
ar
aj
Dep
ar
tm
en
t o
f
C
o
m
p
u
tin
g
T
e
ch
n
o
lo
g
ies,
Facu
lty
o
f
E
n
g
in
e
er
in
g
an
d
T
ec
h
n
o
lo
g
y
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
,
C
h
en
g
alp
attu
Dis
tr
ict,
T
am
il Na
d
u
,
I
n
d
ia
E
m
ail:
s
elv
ar
ap
@
s
r
m
is
t.e
d
u
.
in
1.
I
NT
RO
D
UCT
I
O
N
L
u
n
g
ca
n
ce
r
en
d
u
r
es
o
n
e
o
f
th
e
m
o
s
t
ex
ten
s
iv
e
an
d
d
ea
d
ly
d
is
ea
s
es
th
at
s
p
r
ea
d
s
its
m
o
r
tality
an
n
u
ally
.
E
ar
lier
d
iag
n
o
s
is
o
f
lu
n
g
ca
n
ce
r
an
d
id
en
tif
y
in
g
t
h
e
r
esp
ec
tiv
e
s
tag
es
is
m
o
r
e
c
r
u
cial.
L
u
n
g
t
u
m
o
r
s
tag
es
in
clu
s
iv
ely
r
ef
er
r
e
d
to
as
th
e
tu
m
o
r
,
n
o
d
e,
m
etastas
is
(
T
NM
)
class
if
icatio
n
s
y
s
tem
s
u
ch
as
th
e
s
ize
an
d
ex
ten
t
o
f
th
e
T
,
th
e
p
r
esen
ce
o
f
d
is
tan
t
M
an
d
th
e
in
v
o
lv
em
en
t
o
f
ly
m
p
h
N
[
1
]
–
[
3
]
.
Ad
v
an
ce
d
m
ed
ical
im
ag
in
g
tech
n
o
lo
g
ies
h
a
v
e
a
m
ajo
r
im
p
r
o
v
ed
ab
ilit
y
to
id
e
n
tify
an
d
esti
m
ate
lu
n
g
tu
m
o
r
s
s
u
ch
as
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
(
MRI)
,
c
o
m
p
u
ted
t
o
m
o
g
r
ap
h
y
(
C
T
)
,
an
d
p
o
s
itro
n
em
is
s
io
n
to
m
o
g
r
ap
h
y
(
PET
)
.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
R
esid
u
a
l e
d
g
e
d
en
s
e
en
h
a
n
ce
d
mo
d
u
le
n
etw
o
r
k:
A
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
…
(
P
r
a
b
a
ka
r
a
n
Ja
ya
r
a
ma
n
)
4033
ad
v
an
ce
s
in
ar
tific
ial
in
tellig
e
n
ce
(
AI
)
,
m
ac
h
i
n
e
lear
n
in
g
(
ML
)
,
an
d
d
ee
p
lear
n
in
g
(
DL
)
en
ab
le
m
o
r
e
p
r
ec
is
e
tu
m
o
r
s
tag
in
g
[
4
]
–
[
6
]
.
AI
co
u
ld
tack
le
co
m
p
lex
a
n
d
h
u
g
e
d
atasets
with
p
r
ec
is
e
class
if
icatio
n
.
T
h
e
m
o
s
t
m
o
d
er
n
d
ev
elo
p
m
en
ts
in
im
a
g
in
g
tech
n
o
lo
g
ies
ap
p
ly
A
I
b
ased
m
eth
o
d
s
f
o
r
class
if
y
in
g
t
h
e
tu
m
o
r
s
tag
es.
T
o
av
o
id
h
u
m
an
er
r
o
r
an
d
m
e
d
ical
m
o
d
alities
m
u
s
t
b
e
eq
u
ip
p
ed
with
p
r
ec
is
e
AI
d
r
iv
en
s
o
lu
tio
n
s
.
AI
d
r
iv
en
lu
n
g
tu
m
o
r
class
if
icatio
n
ca
n
also
b
e
ab
le
to
co
n
s
id
er
th
e
in
d
iv
i
d
u
al
p
atien
t’
s
f
ea
tu
r
es
i
n
ca
n
ce
r
ce
ll
s
eg
m
e
n
tatio
n
.
T
h
e
DL
s
u
f
f
er
s
with
th
e
cu
r
s
e
o
f
o
v
er
f
itti
n
g
an
d
r
ed
u
ce
d
th
e
g
en
er
aliza
b
ilit
y
.
Ho
wev
er
,
f
e
atu
r
es f
r
o
m
th
e
C
T
an
d
PET
im
ag
es
h
av
e
v
ar
y
in
g
r
eso
lu
tio
n
s
an
d
p
r
o
p
er
ties
m
a
k
in
g
it
d
if
f
ic
u
lt
to
in
co
r
p
o
r
ate
an
d
ar
r
a
n
g
e
th
e
m
p
r
o
p
er
l
y
[
7
]
.
E
s
p
ec
ially
,
L
u
n
g
tu
m
o
r
is
d
if
f
icu
lt
to
d
etec
t
an
d
ca
teg
o
r
ized
as
t
h
e
tu
m
o
r
s
tag
es
v
ar
ies
with
r
esp
ec
t
to
s
ize,
tex
tu
r
e
an
d
s
h
ap
e.
I
n
r
ea
l
wo
r
ld
clin
ical
s
it
u
atio
n
s
u
p
p
o
r
tin
g
th
e
ac
cu
r
ac
y
o
f
th
e
m
o
d
el
a
n
d
r
eliab
ilit
y
s
p
ec
if
ies
ass
u
r
in
g
th
e
to
u
g
h
n
ess
o
f
th
ese
v
a
r
ia
tio
n
s
.
T
h
e
m
ajo
r
co
n
tr
i
b
u
tio
n
o
f
th
is
wo
r
k
is
m
en
tio
n
ed
as
as f
o
llo
ws:
−
W
e
ad
o
p
t
p
a
r
allel
f
ea
tu
r
e
ex
t
r
ac
to
r
s
n
am
e
d
d
ee
p
r
esid
u
al
co
n
v
o
l
u
tio
n
al
n
etwo
r
k
(
R
C
N)
f
o
r
m
o
d
ality
(
i.e
.
C
T
an
d
PET
)
s
p
ec
if
ic
f
ea
tu
r
e
ex
tr
ac
tio
n
with
s
h
ar
ed
we
ig
h
ts
f
o
r
ex
tr
ac
tin
g
h
ig
h
d
im
e
n
s
io
n
al
s
p
atial
f
ea
tu
r
es.
−
W
e
d
esig
n
UNe
t
++
+
f
o
r
p
er
f
o
r
m
in
g
s
eg
m
en
tatio
n
b
y
e
x
am
in
in
g
h
ig
h
d
im
en
s
io
n
al
s
p
atial
f
ea
tu
r
es
wit
h
b
etter
f
ea
tu
r
e
p
r
o
ce
s
s
in
g
ac
c
u
r
ac
y
.
T
h
e
d
esig
n
ed
UNe
t+
++
co
n
tain
s
co
n
v
o
lu
tio
n
al
b
lo
ck
atten
tio
n
s
eg
m
en
t
(
C
B
AS)
to
r
ed
u
ce
th
e
u
n
wan
ted
co
m
p
u
tatio
n
al
co
m
p
lex
ity
.
T
h
e
a
d
o
p
tin
g
o
f
p
ix
el
lev
el
lear
n
in
g
m
o
d
u
le
(
P2
L
M)
an
d
ed
g
e
lev
el
lear
n
i
n
g
m
o
d
u
le
(
E
2
L
M)
en
h
a
n
ce
th
e
tu
m
o
r
s
eg
m
en
tatio
n
ac
cu
r
ac
y
b
y
ef
f
ec
tiv
ely
p
r
o
ce
s
s
in
g
th
e
m
u
lti s
ca
le
d
ec
o
d
e
d
f
ea
tu
r
es.
T
h
e
r
est
o
f
th
e
s
tu
d
y
is
o
r
g
a
n
ized
as
f
o
llo
ws;
s
ec
tio
n
2
d
em
o
n
s
tr
ates
th
e
r
elate
d
wo
r
k
s
.
Sectio
n
3
em
p
h
ases
th
e
m
ater
ial
an
d
m
eth
o
d
s
n
ee
d
ed
to
p
r
o
p
o
s
ed
r
e
s
ea
r
ch
d
esig
n
.
Sectio
n
4
e
x
p
l
ain
s
th
e
p
r
o
p
o
s
ed
r
esid
u
al
ed
g
e
d
en
s
e
en
h
a
n
c
ed
m
o
d
u
le
n
etwo
r
k
(
R
E
DE
M
-
NE
T
)
m
o
d
el
with
ap
p
r
o
p
r
iate
m
ath
em
atica
l
eq
u
atio
n
s
an
d
d
iag
r
a
m
s
.
Sectio
n
5
im
p
lem
en
ts
th
e
p
r
o
p
o
s
ed
wo
r
k
with
e
x
is
tin
g
wo
r
k
s
.
Sectio
n
6
co
n
cl
u
d
es
th
e
p
r
o
p
o
s
ed
wo
r
k
.
2.
RE
L
AT
E
D
WO
RK
S
Ham
d
i
et
a
l.
[
8
]
h
av
e
d
ev
el
o
p
ed
a
lu
n
g
ca
n
ce
r
class
if
icatio
n
m
eth
o
d
b
y
u
s
in
g
a
m
u
lti
o
u
tp
u
t
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN
)
to
o
l
to
ass
is
t
lu
n
g
ca
n
ce
r
p
atien
t’
s
s
tag
es.
I
t
r
ef
er
s
t
o
th
e
T
NM
s
tag
in
g
m
eth
o
d
a
n
d
h
is
to
lo
g
ic
s
u
b
ty
p
es
class
if
icatio
n
.
Fu
r
th
er
m
o
r
e
,
VGG
-
1
6
n
etwo
r
k
h
as
b
ee
n
in
co
r
p
o
r
ated
with
PET
/C
T
im
ag
es
to
ex
tr
ac
t
r
elev
an
t
f
ea
tu
r
es
f
r
o
m
im
ag
es
.
Acc
o
r
d
in
g
to
Kasin
ath
an
a
n
d
J
ay
ak
u
m
ar
[
9
]
,
a
clo
u
d
-
b
ased
lu
n
g
tu
m
o
r
d
ete
cto
r
an
d
s
tag
e
class
if
ier
(
c
lo
u
d
-
L
T
DSC
)
m
eth
o
d
wer
e
p
r
o
p
o
s
ed
to
class
if
y
an
d
v
alid
ate
lu
n
g
tu
m
o
r
s
tag
es b
y
u
tili
zin
g
DNN
an
d
clo
u
d
-
b
ase
d
d
ata
co
llectio
n
.
T
h
is
m
eth
o
d
was v
alid
ated
with
th
e
b
en
c
h
m
ar
k
lu
n
g
im
a
g
e
d
at
ab
ase
co
n
s
o
r
tiu
m
-
im
ag
e
d
ata
b
ase
r
eso
u
r
ce
in
itiativ
e
(
L
I
DC
-
I
DR
I
)
d
ataset
an
d
C
T
d
ig
ital im
ag
in
g
an
d
co
m
m
u
n
icatio
n
s
in
m
ed
icin
e
(
DI
C
OM
)
im
ag
es.
T
h
e
r
esear
ch
in
[
1
0
]
,
[
1
1
]
h
a
v
e
p
r
esen
ted
a
f
r
am
ewo
r
k
th
at
au
to
m
atica
lly
lo
ca
lizes
lu
n
g
c
an
ce
r
f
r
o
m
PET
/C
T
im
ag
es.
R
o
s
e
et
a
l.
[
1
2
]
p
r
o
p
o
s
ed
a
f
r
am
ewo
r
k
f
o
r
th
e
ca
n
ce
r
d
etec
tio
n
with
f
u
z
zy
C
-
m
ea
n
s
(
FC
M)
.
Xian
g
et
a
l.
[
1
3
]
p
r
o
p
o
s
ed
a
m
o
d
ality
-
s
p
ec
if
ic
s
eg
m
en
tati
o
n
n
etwo
r
k
(
Mo
SNet)
tech
n
i
q
u
e
f
o
r
lu
n
g
t
u
m
o
r
s
eg
m
en
tatio
n
to
y
ield
a
m
o
d
al
ity
-
s
p
ec
if
ic
m
ap
.
Fu
et
a
l.
[
1
4
]
p
r
o
p
o
s
ed
a
lu
n
g
tu
m
o
r
s
eg
m
e
n
tatio
n
m
o
d
el
with
m
u
ltimo
d
al
s
p
atial
atten
tio
n
m
o
d
u
le
(
MSAM
)
.
Xie
et
a
l.
[
1
5
]
p
r
o
p
o
s
ed
a
n
o
v
el
m
et
h
o
d
t
o
id
en
tif
y
th
e
p
r
eo
p
e
r
ativ
e
ly
m
p
h
n
o
d
e
s
tag
in
g
in
n
o
n
-
s
m
all
ce
ll
lu
n
g
ca
n
ce
r
.
Mo
r
eo
v
er
,
r
etr
o
s
p
ec
tiv
e
ex
am
in
atio
n
o
f
2
6
3
ab
n
o
r
m
ally
v
er
if
ied
ly
m
p
h
n
o
d
es f
r
o
m
1
2
4
non
-
s
m
all
ce
ll lu
n
g
ca
n
ce
r
(
NSC
L
C
)
p
atien
ts
was p
er
f
o
r
m
ed
.
X
i
a
a
n
d
Z
h
a
n
g
[
1
6
]
p
r
o
p
o
s
ed
a
n
o
v
e
l
DL
-
b
as
e
d
g
r
a
p
h
m
o
d
e
l
f
o
r
t
u
m
o
r
s
e
g
m
e
n
t
at
i
o
n
w
a
s
.
T
h
ei
r
m
e
t
h
o
d
h
a
s
e
x
p
l
o
i
t
e
d
t
h
e
C
T
’
s
s
p
a
ti
a
l
r
es
o
l
u
t
i
o
n
a
n
d
PE
T
’
s
h
i
g
h
e
r
c
o
n
t
r
a
s
t
f
o
r
m
u
l
t
i
-
s
c
a
l
e
f
u
s
i
o
n
a
n
d
co
-
s
e
g
m
e
n
t
a
ti
o
n
.
N
a
w
r
e
e
n
e
t
a
l
.
[
1
7
]
p
r
o
p
o
s
e
d
a
n
o
v
e
l
m
e
t
h
o
d
f
o
r
p
r
e
-
p
r
o
c
e
s
s
i
n
g
w
it
h
i
m
a
g
e
e
n
h
a
n
c
e
m
e
n
t
a
n
d
s
m
o
o
t
h
i
n
g
.
R
e
h
m
a
n
e
t
a
l
.
[
1
8
]
d
e
m
o
n
s
t
r
a
t
e
d
t
h
e
i
d
e
a
o
f
f
e
a
t
u
r
e
f
u
s
i
o
n
w
i
t
h
p
a
t
c
h
b
a
s
e
lo
c
a
l
b
i
n
a
r
y
p
a
t
t
e
r
n
(
L
B
P
)
a
n
d
d
i
s
c
r
et
e
c
o
s
i
n
e
t
r
a
n
s
f
o
r
m
.
Y
a
d
a
v
e
t
a
l
.
[
1
9
]
p
r
o
p
o
s
e
d
a
f
r
a
m
e
w
o
r
k
f
o
r
c
h
e
s
t
C
T
a
n
d
X
-
r
a
y
i
m
a
g
e
s
w
i
t
h
g
e
n
e
r
a
ti
v
e
a
d
v
e
r
s
a
r
ia
l
n
e
tw
o
r
k
(
GAN
)
.
R
az
a
et
a
l.
[
2
0
]
p
r
o
p
o
s
ed
L
u
n
g
-
E
f
f
Net
f
o
r
lu
n
g
ca
n
ce
r
b
y
u
tili
zin
g
a
E
f
f
icien
tNet
f
r
o
m
C
T
-
s
ca
n
im
ag
es.
T
h
is
m
eth
o
d
h
as
ad
d
i
tio
n
al
to
p
lay
er
s
f
o
r
th
e
class
if
icatio
n
h
ea
d
an
d
it
was
ev
alu
ated
b
y
u
s
in
g
f
iv
e
E
f
f
icien
tNet
v
ar
iatio
n
s
(
B
0
-
B
4
)
.
Mo
r
e
o
v
er
,
it wa
s
ex
p
er
im
e
n
ted
o
n
t
h
e
I
Q
-
OT
H/NC
C
D
b
en
ch
m
ar
k
d
ataset
to
class
if
y
lu
n
g
ca
n
ce
r
as
b
en
ig
n
o
r
m
alig
n
an
t.
Ven
k
ates
h
et
a
l.
[
2
1
]
p
r
o
p
o
s
ed
a
l
u
n
g
ca
n
ce
r
lesi
o
n
s
id
en
tific
atio
n
m
et
h
o
d
with
Ot
s
u
th
r
esh
o
ld
in
g
a
n
d
C
NN
b
ased
cu
c
k
o
o
s
ea
r
ch
alg
o
r
ith
m
.
T
h
is
f
r
am
ewo
r
k
was
v
alid
ated
with
th
e
s
ca
lin
g
,
r
o
t
atio
n
an
d
co
n
tr
ast
m
o
d
if
icatio
n
o
f
th
e
im
ag
es
tak
en
f
r
o
m
L
I
DC
-
I
DR
I
d
atab
ase
[
2
2
]
.
Far
u
q
u
i
et
a
l.
[
2
3
]
p
r
o
p
o
s
ed
a
h
y
b
r
id
d
eep
-
C
NN
m
o
d
el
n
am
ed
L
u
n
g
Net
with
2
2
-
lay
er
h
y
b
r
id
d
ee
p
-
C
NN.
T
h
ey
h
a
v
e
tr
ain
e
d
th
e
m
o
d
el
with
C
T
s
ca
n
s
an
d
wea
r
ab
le
s
en
s
o
r
-
b
ased
MI
o
T
d
at
a.
Naz
ir
et
a
l.
[
2
4
]
p
r
o
p
o
s
ed
a
lu
n
g
s
eg
m
e
n
tatio
n
f
r
am
ewo
r
k
with
L
P
d
ec
o
m
p
o
s
itio
n
an
d
ad
ap
tiv
e
s
p
ar
s
e
r
e
p
r
esen
tatio
n
(
ASR
)
an
d
v
alid
ated
with
th
e
im
ag
es
tak
en
f
r
o
m
L
I
DC
-
I
DR
I
d
atab
a
s
e.
Ash
waty
et
a
l.
[
2
5
]
p
r
o
p
o
s
ed
a
m
o
d
el
to
d
etec
t
lu
n
g
tu
m
o
r
u
s
in
g
Nan
o
-
s
eg
m
en
ted
C
T
im
ag
e.
T
h
is
m
eth
o
d
was
en
h
a
n
ce
d
with
G
ab
o
r
f
ilter
a
n
d
c
o
lo
r
-
b
ased
h
i
s
to
g
r
am
eq
u
aliza
tio
n
tec
h
n
iq
u
es.
Mo
r
eo
v
er
,
th
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
3
2
-
4
0
4
2
4034
lu
n
g
ca
n
ce
r
im
ag
es
wer
e
s
eg
m
en
ted
b
y
u
tili
zin
g
th
e
g
u
a
r
an
teed
co
n
v
e
r
g
en
ce
p
a
r
ticle
s
war
m
o
p
tim
izatio
n
(
GC
PS
O)
alg
o
r
ith
m
.
Ad
d
itio
n
ally
,
th
e
tu
m
o
r
r
eg
i
o
n
s
ar
e
class
if
ied
u
s
in
g
a
g
r
ap
h
ical
u
s
er
in
ter
f
ac
e
to
o
l
an
d
b
ag
-
of
-
v
is
u
al
-
wo
r
d
s
(
B
o
VW
)
a
co
n
v
o
lu
tio
n
al
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
(
C
R
NN)
was
u
tili
ze
d
f
o
r
im
ag
e
class
if
icatio
n
an
d
f
ea
tu
r
e
ex
t
r
ac
tio
n
.
C
r
asta
et
a
l.
[
2
6
]
p
r
o
p
o
s
ed
a
class
if
icatio
n
f
r
am
ewo
r
k
u
s
in
g
c
o
s
in
e
s
ail
f
is
h
o
p
tim
izatio
n
-
b
ased
g
e
n
er
ativ
e
ad
v
er
s
ar
ial
n
etwo
r
k
.
T
h
is
f
r
am
ewo
r
k
m
er
g
es
th
e
s
in
e
co
s
in
e
alg
o
r
ith
m
th
r
o
u
g
h
th
e
s
ailf
is
h
o
p
tim
izer
.
Fu
r
th
er
m
o
r
e,
th
e
p
r
o
ce
s
s
in
clu
d
es
p
r
e
-
p
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
lu
n
g
ca
n
ce
r
d
etec
tio
n
,
s
ev
er
ity
cl
ass
if
icatio
n
an
d
lu
n
g
n
o
d
u
l
e
s
eg
m
en
tatio
n
.
Ad
d
itio
n
all
y
,
C
T
im
ag
es
ar
e
s
eg
m
en
ted
to
d
etec
t a
b
n
o
r
m
al
r
eg
io
n
s
.
3.
M
AT
E
R
I
AL
S AN
D
M
E
T
H
O
DS
Fo
r
o
u
r
p
r
o
p
o
s
ed
r
esear
ch
,
w
e
r
etr
o
s
p
ec
tiv
ely
co
llect
(
n
=
1
2
0
)
s
tag
e
I
lu
n
g
tu
m
o
r
p
atien
ts
h
av
e
a
f
av
o
r
a
b
le
d
iag
n
o
s
is
wh
en
co
m
p
ar
ed
to
th
e
p
atien
ts
with
s
tag
e
I
I
an
d
I
I
I
r
esp
ec
tiv
ely
.
T
o
b
e
clea
r
er
,
it
is
d
if
f
icu
lt
to
v
ictim
ize
th
e
s
tag
e
I
I
an
d
I
I
I
p
atien
ts
with
g
o
o
d
an
d
wo
r
s
e
p
r
o
g
n
o
s
is
.
So
th
at
th
is
s
tu
d
y
f
ir
m
ly
f
o
cu
s
ed
o
n
lu
n
g
tu
m
o
r
p
atien
t
s
with
s
tag
e
I
I
an
d
I
I
I
r
esp
ec
ti
v
ely
.
−
Patien
t
d
em
o
g
r
ap
h
ics:
f
o
r
o
u
r
r
esear
ch
,
we
u
tili
ze
ab
o
u
t
1
4
0
lu
n
g
ca
n
ce
r
p
atien
ts
with
s
t
ag
e
I
I
an
d
I
I
I
wer
e
ad
o
p
ted
f
r
o
m
th
e
C
en
t
r
e
Ho
s
p
italier
Un
iv
er
s
itair
e
(
C
HU)
,
Fra
n
ce
in
wh
ich
th
e
p
atien
ts
ar
e
s
u
b
jecte
d
with
cu
r
ativ
e
ch
em
o
r
ad
io
th
e
r
ap
y
.
T
h
o
s
e
p
atien
ts
’
d
etails
wer
e
co
llected
b
etwe
en
th
e
y
ea
r
s
2
0
1
7
to
2
0
1
9
r
esp
ec
tiv
ely
.
T
h
e
in
clu
s
io
n
cr
iter
ia
in
clu
d
e
p
a
tien
ts
with
NSC
L
C
with
b
o
th
PET
an
d
C
T
im
ag
es
an
d
s
tag
e
-
I
I
o
r
I
I
I
with
s
u
b
jecte
d
to
r
a
d
io
th
er
a
p
y
tr
ea
tm
en
t.
T
ab
le
1
s
h
o
ws
th
e
d
em
o
g
r
ap
h
ics
o
f
p
atien
t c
h
ar
ac
ter
is
tics
.
T
ab
le
2
s
h
o
ws th
e
tu
m
o
r
s
tag
in
g
o
f
lu
n
g
ca
n
ce
r
.
−
PET
/C
T
im
ag
e
ac
q
u
is
itio
n
:
a
ll
th
e
co
n
s
id
er
ed
p
atien
ts
m
u
s
t
u
n
d
er
g
o
PET
/C
T
ac
q
u
is
itio
n
b
ef
o
r
e
s
tag
in
g
an
d
tr
ea
tm
en
t
ap
p
r
o
ac
h
.
E
n
tr
e
n
ch
ed
o
n
th
e
clin
ical
r
o
u
tin
e
p
r
o
to
co
l,
a
b
io
g
r
ap
h
with
2
2
.
7
cm
ax
ial
v
iew
was
tak
en
o
n
m
C
T
5
0
T
o
F
.
Af
ter
7
h
o
f
f
asti
n
g
an
d
65
±
7
m
in
u
tes
o
f
3
.
5
MBq
/
kg
o
f
FDG
(
4
2
3
±
98
MBq
,
r
an
g
e
2
2
5
-
7
0
0
MBq
)
PET
/C
T
im
ag
e
ac
q
u
is
itio
n
was star
ted
.
T
h
e
C
T
im
a
g
e
with
n
o
n
-
co
n
tr
ast
en
h
an
ce
d
,
n
o
n
-
r
esp
ir
at
o
r
y
g
ates
wer
e
ac
q
u
ir
e
d
wi
th
4
m
m
th
ick
n
ess
an
d
0
.
942
×
0
.
942
mm
2
in
-
p
lan
e
th
ick
n
ess
u
s
in
g
1
3
0
k
Vp
m
o
d
u
lati
o
n
s
y
s
tem
.
Fu
r
th
er
m
o
r
e,
th
e
PET
im
ag
es
ca
n
b
e
ac
q
u
ir
e
d
b
ased
o
n
th
e
b
ed
p
o
s
itio
n
ar
r
a
n
g
em
en
t
with
3
.
5
m
in
.
T
h
e
ac
q
u
ir
ed
PET
an
d
C
T
im
ag
es
ar
e
r
ec
o
n
s
tr
u
cted
u
s
in
g
s
p
atial
r
eso
lu
tio
n
m
o
d
ellin
g
an
d
tim
e
o
f
f
lig
h
t
with
2
2
s
u
b
s
ets,
f
o
u
r
iter
atio
n
s
,
v
o
x
el
s
ize
o
f
5
×
5
×
5
mm
2
,
an
d
6
m
m
3
D
g
au
s
s
ian
f
ilter
in
g
ap
p
r
o
ac
h
es r
e
s
p
ec
tiv
ely
.
T
ab
le
1
.
Dem
o
g
r
ap
h
ics o
f
p
ati
en
t
C
h
a
r
a
c
t
e
r
i
s
t
i
c
s
#
o
f
P
a
t
i
e
n
t
s
(
%)
Te
st
(
N
=
5
5
)
(
%)
Tr
a
i
n
(
N
=
9
0
)
(
%)
S
t
a
g
e
I
3
3
3
II
4
5
(
3
3
)
1
9
(
3
2
)
2
8
(
2
8
)
III
9
8
(
7
1
)
3
6
(
6
9
)
6
3
(
7
2
)
Tr
e
a
t
me
n
t
C
H
E
7
2
(
5
3
)
2
5
(
4
7
)
5
9
(
6
8
)
R
A
D
7
0
(
5
1
)
3
0
(
5
7
)
3
2
(
3
6
)
A
g
e
M
e
a
n
±
SD
7
3
.
6
±
9
.
4
7
3
.
8
±
12
7
3
.
6
±
9
.
4
R
a
n
g
e
48
-
96
48
-
91
48
-
96
G
e
n
d
e
r
F
e
mal
e
3
4
(
2
5
)
9
(
1
6
)
2
7
(
3
1
)
M
a
l
e
1
0
6
(
7
9
)
4
6
(
8
8
)
6
4
(
7
3
)
T
ab
le
2
.
L
u
n
g
t
u
m
o
r
s
tag
es
P
r
i
mary
t
u
mo
r
(
TU
)
R
e
g
i
o
n
a
l
l
y
m
p
h
n
o
d
e
s
(
LN
)
D
i
st
a
n
t
m
e
t
a
s
t
a
s
i
s (D
M
)
TU
0
–
N
o
t
u
mo
r
TU
1
–
Tu
mo
r
≤
3
c
m
TU
1
a
–
T
u
m
o
r
≤
1
c
m
TU
1
c
–
T
u
m
o
r
>
1
c
m
t
o
≤
2
c
m
TU
2
–
Tu
mo
r
>
2
c
m
t
o
≤
5
c
m
TU
2
a
–
T
u
m
o
r
>
3
c
m
t
o
≤
4
c
m
TU
2
b
–
T
u
m
o
r
>
4
c
m t
o
≤
5
c
m
TU
3
–
Tu
mo
r
>
5
c
m
t
o
≤
7
cm
TU
4
–
Tu
mo
r
>
7
c
m
LN
0
–
N
o
r
e
g
i
o
n
a
l
n
o
d
e
me
t
a
s
t
a
s
i
s
LN
1
–
I
p
si
l
a
t
e
r
a
l
p
e
r
i
b
r
a
n
c
h
i
a
l
/
p
e
r
i
h
i
l
a
r
/
i
n
t
r
a
p
u
l
mo
n
a
r
y
n
o
d
e
s
LN
2
–
I
p
si
l
a
t
e
r
a
l
m
e
d
i
a
s
t
i
n
a
l
o
r
su
b
c
a
r
i
n
a
l
n
o
d
e
s
LN
3
–
C
o
n
t
r
a
l
a
t
e
r
a
l
med
i
a
s
t
i
n
a
l
/
p
e
r
i
h
i
l
a
r
/
su
p
r
a
c
l
a
v
i
c
u
l
a
r
n
o
d
e
s
D
M
0
–
N
o
d
i
st
a
n
t
m
e
t
a
st
a
si
s
D
M
1
–
M
a
l
i
g
n
a
n
t
e
f
f
u
si
o
n
o
r
c
o
n
t
r
a
l
a
t
e
r
a
l
n
o
d
u
l
e
D
M
2
–
D
i
s
t
a
n
t
me
t
a
s
t
a
s
i
s
d
e
t
e
c
t
e
d
4.
RE
D
E
M
N
ET
–
M
O
DE
L
Fig
u
r
e
1
r
ep
r
esen
ts
th
e
ar
ch
it
ec
tu
r
e
o
f
th
e
p
r
o
p
o
s
ed
R
E
DE
M
-
NE
T
.
T
h
e
p
r
o
p
o
s
ed
R
E
DE
M
-
NE
T
is
co
m
p
o
s
ed
o
f
p
ar
allel
f
ea
tu
r
e
ex
tr
ac
to
r
s
n
am
e
d
d
e
n
s
e
r
esid
u
al
co
n
v
o
lu
tio
n
al
n
etwo
r
k
(
D
R
C
N)
f
o
r
ex
tr
ac
tin
g
h
ig
h
d
im
e
n
s
io
n
al
f
ea
tu
r
es
f
r
o
m
b
o
th
t
h
e
PET
an
d
C
T
m
o
d
alities
.
No
te
th
at
th
e
f
ea
tu
r
e
ex
tr
ac
to
r
s
s
h
ar
ed
weig
h
ts
am
o
n
g
th
em
s
elv
es to
ex
am
in
e
th
e
b
is
p
atial
f
ea
tu
r
es.
B
o
th
p
r
o
d
u
ce
h
ig
h
d
im
e
n
s
io
n
al
-
s
p
atial
p
ix
elate
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
R
esid
u
a
l e
d
g
e
d
en
s
e
en
h
a
n
ce
d
mo
d
u
le
n
etw
o
r
k:
A
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
…
(
P
r
a
b
a
ka
r
a
n
Ja
ya
r
a
ma
n
)
4035
f
ea
tu
r
es
an
d
p
r
o
v
id
e
t
h
em
as
in
p
u
t
to
th
e
UN
e
t
+
+
+
.
T
h
e
UN
e
t
+
+
+
ex
tr
ac
ts
m
u
lti
-
lev
el
d
ec
o
d
ed
f
ea
t
u
r
es
an
d
r
eso
lv
e
th
e
r
ed
u
n
d
an
t
co
m
p
u
t
atio
n
p
r
o
b
lem
s
b
y
in
clu
d
in
g
C
B
AS.
T
h
e
o
u
tp
u
t
o
f
UN
e
t
+
+
+
is
d
ep
icted
in
f
o
r
m
o
f
m
u
lti
-
lev
el
f
ea
tu
r
e
m
a
p
s
.
Fro
m
th
e
m
u
lti
-
lev
el
f
ea
tu
r
e
m
ap
s
,
P2
L
M
a
n
d
E
2
L
M
o
u
tp
u
t
th
e
p
r
ec
is
e
lu
n
g
tu
m
o
r
s
eg
m
e
n
tatio
n
r
esu
lt.
Fin
ally
,
th
e
d
etec
ted
t
u
m
o
r
is
th
en
f
ed
to
t
h
e
m
u
lti
class
-
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
MC
-
SVM)
f
o
r
m
u
lti cla
s
s
lu
n
g
tu
m
o
r
s
tag
e
class
if
icatio
n
with
p
ea
k
ac
c
u
r
ac
y
a
n
d
p
r
ec
is
io
n
r
esp
ec
tiv
ely
.
D
RCH
D
RCH
P
E
T
CT
H
D
S
H
D
S
P
E
T
CT
U
N
e
t
+
+
+
P
2L
M
E
2L
M
MC
-
S
V
M
P
ri
m
a
ry T
um
or
Re
gi
on L
ym
ph N
ode
D
i
s
t
a
nt
M
e
t
a
s
t
a
s
i
s
Fig
u
r
e
1
.
Ov
e
r
all
ar
ch
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
R
E
DE
M
-
NE
T
m
o
d
el
4
.
1
.
P
a
ra
llel f
ea
t
ure
ex
t
ra
c
t
o
rs
T
h
e
p
r
e
-
p
r
o
ce
s
s
ed
PET
an
d
C
T
im
ag
e
Im
PET
′′′
,
an
d
Im
CT
′′′
is
p
r
o
v
id
ed
as
an
in
p
u
t
to
th
e
p
a
r
allel
f
ea
tu
r
e
ex
tr
ac
to
r
s
n
am
e
d
DR
C
N
wh
ich
p
r
o
v
id
es
o
u
tp
u
t
as
as
s
o
ciate
d
p
ix
el
-
s
u
p
er
p
ix
els
∀
an
d
h
ig
h
d
im
en
s
io
n
al
s
p
atial
f
ea
tu
r
es
ℵ
ds
.
No
te
th
at,
p
ix
els
an
d
s
u
p
er
-
p
ix
els
in
th
e
m
ed
ical
im
ag
es
ca
n
p
er
f
o
r
m
tr
an
s
f
o
r
m
atio
n
am
o
n
g
ea
c
h
o
t
h
er
∀
∈
ℝ
M
×
N
,
Im
∈
ℝ
M
×
C
,
an
d
SP
∈
ℝ
N
×
C
in
wh
ich
p
ix
els
an
d
s
u
p
er
-
p
ix
els
ar
e
r
ep
r
esen
ted
b
y
M
a
n
d
N
r
esp
ec
tiv
ely
.
Fu
r
th
er
m
o
r
e,
th
e
co
l
o
r
s
ca
lin
g
an
d
p
o
s
itio
n
al
f
ac
t
o
r
s
ar
e
d
e
n
o
ted
b
y
δ
p
o
s
i
,
a
n
d
δ
CSC
.
T
h
e
f
o
r
m
u
latio
n
o
f
δ
p
o
s
i
is
co
m
p
u
ted
in
(
1
)
:
δ
p
o
s
i
=
ℶ
(
N
ω
M
ω
,
N
h
M
h
)
(
1
)
F
r
o
m
(
1
)
,
N
ω
,
N
h
,
M
ω
,
an
d
M
h
d
e
n
o
tes
t
h
e
n
u
m
b
er
o
f
p
ix
els
with
h
eig
h
t
an
d
wid
t
h
r
esp
ec
tiv
ely
.
W
ith
th
e
co
lo
r
s
ca
lin
g
an
d
p
o
s
itio
n
al
f
ac
to
r
s
,
th
e
s
p
atial
co
m
p
ac
tn
ess
an
d
co
lo
r
s
im
ilar
itie
s
ca
n
o
v
er
lo
o
k
th
e
tu
m
o
r
b
o
u
n
d
ar
ies.
T
h
e
d
esig
n
ed
D
R
C
N
en
h
an
ce
s
th
e
lear
n
in
g
ca
p
ab
ilit
y
an
d
ef
f
icac
y
o
f
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
wh
ich
tr
an
s
f
o
r
m
s
th
e
g
iv
en
Im
PET
′′′
o
r
Im
CT
′′′
in
to
two
lev
el
u
p
s
am
p
l
ed
an
d
d
o
wm
s
am
p
led
f
ea
tu
r
e
m
ap
s
r
esp
ec
tiv
ely
.
T
h
e
m
ath
em
ati
ca
l
(
1
)
o
f
th
e
d
en
s
e
co
n
v
o
l
u
tio
n
al
b
lo
ck
s
an
d
s
tack
ed
r
esid
u
al
b
lo
ck
s
ar
e
co
m
p
u
ted
as
(
2
)
to
(
5
)
:
IF
DC
=
PS
+
M
PL
R
↑
4
(
2
)
IF
r
es
1
=
F
e
a
DC
(
3
)
IF
r
es
2
=
F
e
a
DC
⊕
F
e
a
r
es
1
(
4
)
IF
r
es
3
=
(
F
e
a
DC
⊕
F
e
a
r
es
1
)
⊕
F
e
a
r
es
2
(
5
)
Fro
m
(
2
)
-
(
5
)
IF
DC
an
d
F
e
a
DC
d
en
o
tes
th
e
in
p
u
t
an
d
o
u
tp
u
t
f
ea
tu
r
e
m
a
p
s
o
f
th
e
d
en
s
e
co
n
v
o
lu
tio
n
al
b
lo
ck
s
r
esp
ec
tiv
ely
;
IF
r
es
j
an
d
F
e
a
r
es
j
d
en
o
tes
th
e
in
p
u
t
a
n
d
o
u
tp
u
t
f
ea
tu
r
es
m
ap
s
in
th
e
j
-
th
r
esid
u
al
b
lo
ck
s
;
an
d
↑
4
d
en
o
tes th
e
up
-
s
am
p
lin
g
o
p
e
r
atio
n
with
⊕
as a
d
d
itio
n
o
p
er
ati
o
n
o
f
ev
er
y
f
ea
tu
r
e
elem
en
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
3
2
-
4
0
4
2
4036
B
y
en
tr
en
ch
i
n
g
d
en
s
e
co
n
v
o
l
u
tio
n
al
lay
er
s
,
elem
e
n
t
wis
e
ad
d
itio
n
,
a
n
d
s
tack
ed
r
esid
u
al
b
l
o
ck
s
s
u
m
s
th
e
f
ea
tu
r
e
m
ap
s
o
f
ev
e
r
y
d
ep
th
(
i.e
.
6
4
f
ea
tu
r
e
m
a
p
s
)
.
T
h
e
d
ee
p
er
la
y
er
b
etter
will
b
e
s
p
atial
in
f
o
r
m
atio
n
in
th
e
f
ea
tu
r
e
m
ap
s
.
T
o
tally
,
we
h
av
e
u
tili
ze
d
2
0
0
f
ea
tu
r
e
m
ap
s
in
tr
ip
le
le
v
els
an
d
ar
e
am
al
g
am
ated
u
s
in
g
th
e
2
×
2
co
n
v
o
lu
tio
n
al
la
y
er
o
p
er
atio
n
.
T
o
b
e
clea
r
e
r
,
th
e
o
u
tp
u
t
r
esid
u
al
in
f
o
r
m
atio
n
o
f
th
e
f
u
s
ed
with
th
e
Up
s
am
p
le
r
esid
u
al
f
ea
tu
r
e
m
a
p
to
p
r
o
d
u
ce
m
u
lti
s
ca
le
a
n
d
m
u
lti
b
r
a
n
ch
f
ea
tu
r
e
m
ap
s
wi
th
en
r
ich
ed
s
p
atial
in
f
o
r
m
atio
n
.
T
h
e
o
u
tp
u
ts
o
f
th
e
Im
PET
′′′
o
r
Im
CT
′′′
ar
e
h
ig
h
d
im
en
s
io
n
al
s
p
atial
f
ea
tu
r
es
as
HDS
PET
,
a
n
d
HDS
CT
r
esp
ec
tiv
ely
.
4
.
2
.
M
ulti
-
lev
el
deco
ded f
ea
t
ure
ex
t
ra
ct
io
n
T
h
e
o
u
tp
u
t
f
r
o
m
th
e
DR
C
N
HDS
PET
,
a
n
d
HDS
CT
ar
e
p
r
o
v
id
ed
as
an
in
p
u
t
to
th
e
UN
e
t
+
+
+
f
o
r
m
u
lti
-
lev
el
d
ec
o
d
ed
f
ea
t
u
r
e
ex
tr
ac
tio
n
.
T
h
o
s
e
f
ea
tu
r
es
ar
e
p
r
o
v
id
ed
to
ev
e
r
y
en
co
d
e
r
s
tag
e
HDS
En
c
1
,
HDS
En
c
2
,
HDS
En
c
3
,
a
n
d
HDS
En
c
4
an
d
ar
e
am
al
g
am
ated
a
n
d
p
a
s
s
ed
to
th
e
C
B
AS
in
th
e
d
ec
o
d
er
b
lo
ck
HDS
D
ec
1
,
HDS
D
ec
2
,
HDS
D
ec
3
,
a
n
d
HDS
D
ec
4
b
y
p
er
f
o
r
m
i
n
g
2
×
2
co
n
v
o
l
u
tio
n
al
an
d
b
atch
n
o
r
m
aliza
tio
n
o
p
e
r
atio
n
r
esp
ec
tiv
ely
.
T
h
e
m
ajo
r
a
d
v
a
n
tag
e
o
f
u
tili
zin
g
UN
e
t
+
+
+
o
v
e
r
co
n
v
e
n
tio
n
al
UNe
t
is
th
at
th
e
d
esig
n
ed
m
o
d
el
u
tili
ze
s
o
n
ly
less
er
p
ar
am
eter
s
f
o
r
m
u
lti s
ca
le
d
ec
o
d
ed
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
f
u
s
io
n
r
esp
e
ct
iv
ely
.
T
h
e
s
k
ip
co
n
n
ec
tio
n
s
i
n
UN
e
t
+
+
+
en
h
a
n
ce
m
u
lti
-
s
ca
le
f
ea
tu
r
e
ex
tr
ac
t
io
n
b
u
t
ca
n
ca
u
s
e
r
ed
u
n
d
a
n
cy
.
T
o
ad
d
r
ess
th
is
,
f
ea
tu
r
es
f
r
o
m
HDS
D
ec
3
ar
e
p
r
o
ce
s
s
ed
th
r
o
u
g
h
HDS
En
c
1
an
d
HDS
En
c
2
f
r
o
m
d
if
f
er
en
t
m
ax
-
p
o
o
lin
g
lay
er
s
.
HDS
En
c
3
,
HDS
D
ec
4
,
a
n
d
HDS
D
ec
5
ar
e
r
e
f
in
ed
u
s
in
g
2
×2
c
o
n
v
o
lu
tio
n
s
w
ith
s
ig
m
o
id
ac
tiv
atio
n
an
d
bi
-
lin
ea
r
u
p
s
am
p
lin
g
,
e
n
ab
lin
g
ef
f
icien
t sem
an
tic
f
ea
tu
r
e
lear
n
in
g
with
r
e
d
u
ce
d
c
o
m
p
u
tatio
n
al
co
m
p
lex
ity
.
T
h
e
f
o
r
m
u
latio
n
o
f
s
k
ip
c
o
n
n
ec
ted
m
u
ltis
ca
le
UN
e
t
+
+
+
is
p
r
o
v
id
e
d
as
f
o
llo
ws;
a
s
s
u
m
e
th
at
j
b
e
t
h
e
p
r
esen
t
en
c
o
d
er
an
d
d
ec
o
d
er
lay
er
r
esp
ec
tiv
el
y
with
M
n
u
m
b
er
if
o
v
e
r
all
lay
er
s
.
T
h
e
f
ea
t
u
r
e
m
a
p
s
ar
r
a
n
g
ed
in
s
tack
ed
f
o
r
m
at
ar
e
d
e
n
o
ted
b
y
HDS
D
ec
j
th
at
ca
n
b
e
ca
lcu
lated
as
(
6
)
:
HDS
D
ec
j
=
{
B
(
CBN
[
(
B
N
R
(
HDS
En
c
j
)
,
UP
(
HDS
D
ec
k
)
k
=
j
+
1
M
−
2
]
)
)
,
j
=
1
B
(
CBN
[
(
B
N
R
(
HDS
En
c
j
)
,
UP
(
HDS
D
ec
k
)
k
=
j
+
1
,
DS
(
HDS
D
ec
k
)
i
=
j
−
2
]
)
)
,
j
>
1
,
>
0
(
6
)
F
r
o
m
(
6
)
,
t
h
e
C
B
AS
is
d
en
o
t
ed
b
y
B
(
.
)
wh
ich
is
s
u
cc
ee
d
e
d
b
y
th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
.
T
h
e
C
B
N
(
)
,
an
d
B
NR
(
)
i
s
d
en
o
ted
b
y
co
n
v
o
lu
tio
n
b
atch
n
o
r
m
aliza
tio
n
s
et
an
d
R
eL
U
b
atch
n
o
r
m
aliza
tio
n
s
et
r
esp
ec
tiv
ely
,
UP
(
)
,
an
d
DS
(
)
d
e
n
o
tes
th
e
u
p
s
am
p
lin
g
a
n
d
d
o
wn
s
am
p
lin
g
o
p
er
atio
n
r
esp
ec
tiv
ely
.
T
h
e
ad
o
p
tio
n
o
f
C
B
AS
with
b
o
th
th
e
s
p
atial
an
d
ch
an
n
el
atten
tio
n
en
h
an
c
es
th
e
co
n
tex
tu
al
in
f
o
r
m
atio
n
in
m
u
lti
s
ca
le
f
ash
io
n
.
I
n
o
u
r
d
esig
n
,
d
ec
o
d
er
lay
er
attain
s
f
ea
tu
r
e
m
ap
M
SF
∈
R
W
×
H
×
C
in
wh
ich
it
is
p
r
o
v
i
d
ed
as
an
in
p
u
t
to
th
e
C
B
AS.
Af
ter
m
ath
,
th
e
ch
an
n
el
en
tr
e
n
ch
ed
f
ea
tu
r
e
m
a
p
N
c
∈
R
C
×
1
×
1
an
d
s
p
atial
en
tr
en
ch
ed
f
ea
tu
r
e
m
ap
N
s
∈
R
1
×
H
×
W
.
T
h
e
f
in
al
f
ea
tu
r
e
m
ap
o
u
tp
u
t
fe
a
′′
is
co
m
p
u
ted
as
(
7
)
an
d
(
8
)
:
fe
a
′
=
N
c
(
fe
a
)
⊗
fe
a
(
7
)
fe
a
′′
=
N
s
(
fe
a
′
)
⊗
fe
a
′
(
8
)
T
he
r
ef
in
e
d
m
u
ltis
ca
le
f
ea
tu
r
e
m
ap
fe
a
′′
h
o
ld
s
th
e
m
u
lti s
ca
le
s
em
an
tic
in
f
o
r
m
atio
n
o
f
th
e
lu
n
g
tu
m
o
r
s
.
4
.
3
.
P
i
x
el
-
edg
e
lev
el
lea
rning
T
h
e
ex
tr
ac
te
d
fe
a
′′
is
th
en
p
ass
ed
s
im
u
ltan
eo
u
s
ly
to
th
e
P2
L
M
an
d
E
2
L
M
f
o
r
h
ig
h
ly
p
r
éc
is
ed
lu
n
g
tu
m
o
r
s
eg
m
en
tatio
n
.
T
h
e
P2
L
M
is
co
m
p
o
s
ed
o
f
s
er
ies
o
f
c
o
n
v
o
lu
ti
o
n
al
lay
er
s
wh
ich
o
u
tp
u
ts
th
e
p
ix
el
wis
e
class
if
icatio
n
m
ap
.
T
h
e
fe
a
′′
(
PET
)
is
p
ass
ed
to
th
e
s
er
ies
o
f
3
×
3
co
n
v
o
lu
tio
n
lay
e
r
s
in
wh
ich
t
h
e
f
in
al
co
n
v
o
l
u
tio
n
al
lay
er
c
on
v
op
ca
n
b
e
f
o
r
m
u
lated
as
(
9
)
:
L
=
c
on
v
op
(
fe
a
′′
(
PET
)
)
(
9
)
F
r
o
m
(
9
)
,
L
∈
R
H
×
W
×
C
d
e
n
o
t
e
s
t
h
e
l
o
g
it
m
ap
h
e
r
e
C
d
e
n
o
t
e
s
t
h
e
n
u
m
b
e
r
o
f
c
l
a
s
s
es
.
W
e
a
p
p
l
y
S
o
f
t
M
a
x
f
u
n
c
t
i
o
n
t
o
f
e
a
t
u
r
e
m
a
p
c
h
a
n
n
e
l
d
i
m
e
n
s
io
n
t
o
a
t
t
a
i
n
t
h
e
p
r
o
b
a
b
i
li
t
y
m
a
p
Pr
.
T
h
e
f
o
r
m
u
l
a
t
i
o
n
o
f
Pr
i
s
d
e
n
o
t
ed
a
s
(
1
0
)
:
Pr
j
,
i
,
c
=
e
L
j
,
i
,
c
∑
e
L
j
,
i
,
c
′
C
′
c
′
=
1
(
1
0
)
T
he
lo
s
s
f
u
n
ctio
n
f
o
r
p
ix
el
wis
e
lear
n
in
g
u
tili
ze
d
is
cr
o
s
s
en
tr
o
p
y
lo
s
s
an
d
ca
n
b
e
a
u
th
o
r
ized
as
(
1
1
)
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
R
esid
u
a
l e
d
g
e
d
en
s
e
en
h
a
n
ce
d
mo
d
u
le
n
etw
o
r
k:
A
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
…
(
P
r
a
b
a
ka
r
a
n
Ja
ya
r
a
ma
n
)
4037
Lo
=
1
H
′
W
′
∑
∑
∑
Gr
j
,
i
,
c
C
′
c
=
1
W
′
i
=
1
H
′
j
=
1
l
og
(
Pr
j
,
i
,
c
)
(
1
1
)
F
r
o
m
(
1
1
)
,
Gr
∈
R
H
′
×
W
′
×
C
′
d
en
o
tes th
e
g
r
o
u
n
d
tr
u
th
p
ix
el
lev
el
p
r
o
b
a
b
ilit
y
m
ap
.
T
h
e
ed
g
e
f
ea
tu
r
es
ca
n
b
e
e
x
tr
ac
ted
f
r
o
m
t
h
e
fe
a
′′
(
CT
)
f
ea
tu
r
e
m
a
p
u
s
in
g
th
e
g
en
er
al
ed
g
e
d
etec
tio
n
o
p
er
ato
r
Υ
as
s
h
o
wn
in
(
1
2
)
:
Ed
=
Υ
(
fe
a
′′
(
CT
)
)
(
1
2
)
F
r
o
m
(
1
2
)
,
Ed
∈
R
H
×
W
f
r
o
m
th
e
fe
a
′′
(
CT
)
in
p
u
t.
Af
ter
m
ath
,
f
ea
t
u
r
e
e
x
tr
ac
tio
n
ca
n
b
e
p
er
f
o
r
m
e
d
b
ased
o
n
th
e
co
n
te
x
tu
al
in
f
o
r
m
ati
o
n
a
r
o
u
n
d
th
e
im
a
g
e
e
d
g
es.
Ass
u
m
e
th
at
c
o
n
v
o
lu
tio
n
al
lay
er
s
ar
e
ad
o
p
ted
f
o
r
f
ea
tu
r
e
m
ap
as
(
1
3
)
:
F
e
a
k
=
c
on
k
(
F
e
a
k
−
1
)
,
k
∈
{
1
,
2
,
…
,
K
}
(
1
3
)
F
r
o
m
(
1
3
)
,
F
e
a
0
=
fe
a
′′
(
CT
)
as
th
e
in
p
u
t
i
m
ag
e,
an
d
F
e
a
k
∈
R
H
k
×
W
k
×
D
k
is
th
e
k
-
th
lay
er
f
ea
tu
r
e
m
a
p
.
Ad
o
p
ted
f
ea
t
u
r
e
m
a
p
is
th
en
p
r
o
v
id
ed
o
n
c
o
n
v
o
lu
tio
n
al
lay
e
r
f
o
r
e
d
g
e
lev
el
class
if
icatio
n
as
(
1
4
)
:
ℑ
=
c
on
ed
g
(
fe
a
i
′
)
(
1
4
)
F
r
o
m
(
1
4
)
,
ℑ
d
e
n
o
tes
lo
g
it
e
d
g
e
m
ap
with
c
′
as
th
e
ed
g
e
class
es
.
I
n
s
im
ilar
m
an
n
e
r
to
(
9
)
,
s
o
f
t
m
ax
o
p
er
atio
n
also
p
er
f
o
r
m
e
d
f
o
r
g
e
n
er
atin
g
ed
g
e
lev
el
f
ea
tu
r
e
m
ap
.
B
y
co
m
b
in
in
g
th
e
P2
L
M
an
d
E
2
L
M,
we
o
b
tain
th
e
s
eg
m
en
ted
o
u
t
p
u
t a
n
d
ca
n
b
e
f
o
r
m
u
lated
a
s
(
1
5
)
:
Se
g
LC
=
L
(
fe
a
′′
(
PET
)
)
⊕
ℑ
(
fe
a
′′
(
CT
)
)
(
1
5
)
4
.
4
.
M
a
chine le
a
rning
ba
s
ed
lun
g
ca
ncer
s
t
a
g
e
cla
s
s
if
ica
t
io
n
Fro
m
th
e
Se
g
LC
we
d
er
iv
e
w
-
class
es
with
w
-
b
in
ar
y
class
if
ier
s
ar
e
tr
ain
ed
i
n
wh
ich
ev
er
y
b
i
n
ar
y
class
if
ier
s
ar
e
tr
ain
ed
to
d
etac
h
ed
th
e
j
-
th
class
f
r
o
m
o
t
h
e
r
class
es.
Fo
r
en
ab
lin
g
MC
-
S
VM
,
th
is
r
esear
ch
u
tili
ze
d
o
n
e
Vs
r
est
(
OV
R
)
ap
p
r
o
ac
h
wh
ich
s
p
lits
th
e
m
u
ltip
le
b
in
ar
y
class
if
ier
s
.
T
o
b
e
m
o
r
e
d
is
tin
ctiv
e,
th
e
tr
ain
in
g
p
r
o
ce
s
s
with
ev
er
y
w
-
th
b
in
ar
y
class
if
ier
with
j
-
th
class
ca
n
b
e
f
o
r
m
u
lated
as
(
1
6
)
:
min
i
we
j
,
bi
j
,
μ
j
1
2
‖
we
j
‖
2
+
RP
∑
μ
ji
n
j
=
1
(
1
6
)
Su
b
ject
to
th
e
co
n
s
tr
ain
t in
(
1
7
)
:
Ⅎ
ji
(
we
j
.
ð
i
+
bi
j
)
≥
1
−
μ
ji
,
∀
i
=
1
,
2
,
…
,
n
(
1
7
)
W
h
er
e
th
e
lab
el
f
u
n
ctio
n
is
d
e
f
in
ed
in
(
1
8
)
:
Ⅎ
ji
=
{
−
1
if
Ⅎ
i
=
j
1
if
Ⅎ
i
≠
j
(
1
8
)
B
ased
o
n
th
e
f
o
r
m
u
latio
n
,
th
e
MC
-
SVM
clas
s
if
ies
th
e
Se
g
LC
t
u
m
o
r
in
to
th
r
ee
s
tag
es
as
p
r
im
ar
y
tu
m
o
r
,
r
eg
io
n
al
ly
m
p
h
n
o
d
e,
an
d
d
is
tan
t m
etastas
is
.
5.
E
XP
E
R
I
M
E
N
T
A
L
E
VA
L
U
AT
I
O
N
5
.
1
.
I
m
ple
m
ent
a
t
io
n det
a
ils
W
e
u
tili
ze
d
th
e
p
y
th
o
n
T
e
n
s
o
r
Flo
w
an
d
Ker
as
p
ac
k
ag
es
to
em
p
lo
y
o
u
r
a
p
p
r
o
ac
h
.
W
e
also
im
p
lem
en
ted
a
n
NVI
DI
A
T
esla
T
4
GPU
wh
ich
Go
o
g
le
C
o
lab
o
f
f
e
r
s
f
o
r
tr
ain
in
g
a
n
d
co
n
tr
o
llin
g
th
e
class
if
ier
.
W
e
u
til
ized
m
in
i
b
atch
es
o
f
Size
1
6
to
tr
ain
th
e
n
etwo
r
k
f
o
r
1
0
0
ep
o
c
h
s
with
th
e
p
ast
s
to
p
p
in
g
co
n
d
itio
n
s
s
et
to
1
0
ep
o
c
h
s
to
av
o
id
o
v
e
r
f
itti
n
g
.
W
e
co
m
m
o
n
ly
ad
ju
s
t
th
e
am
o
u
n
t
o
f
th
e
ex
is
tin
g
wo
r
k
an
d
b
eg
in
th
e
t
r
ain
in
g
p
r
o
ce
s
s
f
r
o
m
s
cr
atch
.
W
e
also
em
p
lo
y
e
d
th
e
Ad
am
o
p
tim
izer
to
tr
ain
th
e
n
etwo
r
k
a
b
o
v
e
to
b
elo
w.
E
s
p
ec
ially
,
we
p
lace
th
e
in
itial
lear
n
in
g
r
ate
to
0
.
0
0
0
1
an
d
th
e
o
p
tim
izer
’
s
p
ar
am
eter
s
to
0
.
9
a
n
d
0
.
9
9
9
,
if
th
e
r
e
is
n
o
v
alid
atio
n
lo
s
s
en
h
an
ce
m
e
n
t
is
n
o
ticed
f
o
r
f
i
v
e
co
n
s
ec
u
tiv
e
ep
o
c
h
s
.
W
e
also
f
in
e
-
tu
n
ed
th
e
r
ec
en
t m
o
d
el
f
o
r
f
iv
e
e
p
o
c
h
s
at
a
lear
n
in
g
r
ate
to
ad
v
er
s
ely
d
am
ag
e
t
h
e
weig
h
ts
o
f
t
h
e
m
o
d
el.
5
.
2
.
E
v
a
lua
t
i
o
n
m
et
rics
T
h
e
em
p
l
o
y
ed
DL
tech
n
i
q
u
es
ar
e
ev
alu
ated
in
t
h
e
p
r
o
v
is
io
n
s
o
f
f
iv
e
m
etr
ics
f
o
r
s
ep
ar
ate
p
lan
e
s
u
ch
as
a
cc
u
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
i
n
ter
s
ec
tio
n
o
v
er
u
n
i
o
n
(
I
o
U)
,
an
d
a
r
ea
u
n
d
e
r
cu
r
v
e
(
A
U
C
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
3
2
-
4
0
4
2
4038
−
Acc
u
r
ac
y
:
ac
cu
r
ac
y
is
d
ef
in
ed
as
th
e
r
atio
o
f
p
r
o
p
er
ly
d
etec
ted
th
e
v
ar
io
u
s
p
ix
els
an
d
th
e
ea
s
iest
m
etr
ic
o
n
th
e
f
i
v
e
m
etr
ics.
−
Pre
cisi
o
n
:
p
r
ec
is
io
n
is
s
p
ec
if
ie
d
as
th
e
m
ea
s
u
r
e
o
f
th
e
p
r
o
p
er
ly
d
etec
ted
m
alig
n
an
t
p
ix
els
t
o
th
e
d
if
f
er
e
n
t
p
ix
els
ca
teg
o
r
ized
as
m
alig
n
a
n
t
in
th
e
p
lan
e
an
d
t
h
u
s
it
d
is
p
lay
s
th
e
b
etter
m
o
d
el
is
wh
ile
th
e
r
esu
lts
ar
e
p
o
s
itiv
e.
−
R
ec
all:
r
ec
all
is
th
e
p
iece
o
f
t
h
e
m
alig
n
an
t
p
ix
els
in
th
e
b
o
t
to
m
tr
u
th
w
h
ich
wer
e
p
r
o
p
er
l
y
d
etec
ted
a
n
d
class
if
ied
as
to
all
m
alig
n
an
t
p
ix
els
in
th
e
d
ata
s
et.
A
le
s
s
r
e
ca
ll
s
ig
n
if
ies
a
h
u
g
e
n
u
m
b
er
o
f
f
ak
e
alar
m
s
,
y
et
a
less
p
r
ec
is
io
n
v
alu
e
s
p
ec
if
ies a
h
u
g
e
n
u
m
b
er
o
f
f
ak
e
al
ar
m
s
.
−
F1
-
s
co
r
e:
F1
-
s
co
r
e
is
d
ef
in
e
d
as th
e
co
n
s
o
n
an
t
m
ea
n
o
f
th
e
p
r
ec
is
io
n
an
d
r
ec
all.
−
I
o
U
:
I
o
U
is
d
ef
i
n
ed
as
th
e
in
t
er
s
ec
tio
n
m
ea
s
u
r
e
o
f
th
e
g
r
o
u
n
d
tr
u
th
u
n
io
n
a
n
d
d
etec
ted
p
l
ac
es
an
d
is
th
e
ess
en
tial m
etr
ic
o
n
s
ev
er
al
o
b
j
ec
t c
lass
if
icatio
n
an
d
s
eg
m
en
t
atio
n
is
s
u
es.
−
AUC
:
A
UC
is
s
p
ec
if
ied
as
th
e
p
lace
b
elo
w
th
e
R
OC
cu
r
v
e
.
R
OC
cu
r
v
e
h
as
th
e
r
ec
all
o
f
th
e
d
esig
n
o
n
th
e
y
-
ax
is
an
d
f
alse
p
o
s
itiv
e
p
er
ce
n
tag
e
+
f
alse
p
o
s
itiv
e
ex
am
p
les
(
FPE
)
.
FP
E
tr
u
e
n
eg
a
tiv
e
ex
am
p
les
(
T
NE
)
o
n
t
h
e
x
-
ax
is
.
AUC
i
s
a
p
er
ce
n
tag
e
o
f
th
e
m
o
d
el’
s
ef
f
ec
tiv
en
ess
in
an
in
cr
ea
s
e
d
u
n
b
alan
ce
d
o
u
tlin
e
an
d
will b
e
ev
alu
ated
a
s
th
e
v
ital b
elo
w
th
e
R
O
C
cu
r
v
e.
T
h
e
m
en
tio
n
ed
m
etr
ics ar
e
ev
alu
ated
f
o
r
g
iv
en
p
la
n
e
as
in
(
1
9
)
to
(
2
2
)
:
A
c
c
ura
c
y
=
T
PE
+
T
N
E
T
PE
+
T
N
E
+
F
PE
+
F
N
E
(
1
9
)
Pr
e
c
isio
n
=
T
PE
T
PE
+
F
PE
(
2
0
)
R
e
c
a
l
l
=
T
PE
T
PE
+
F
N
E
(
2
1
)
F1
−
s
c
or
e
=
T
PE
T
PE
+
1
2
(
F
PE
+
F
N
E
)
(
2
2
)
H
er
e
t
r
u
e
p
o
s
i
ti
v
e
e
x
am
p
l
es
(
T
PE
)
,
T
N
E
,
FP
E
,
a
n
d
f
alse
n
e
g
a
tiv
e
e
x
a
m
p
les
(
FN
E
)
s
i
g
n
i
f
i
es th
e
am
o
u
n
t
o
f
t
r
u
e
p
o
s
iti
v
e
,
t
r
u
e
n
e
g
ati
v
e
,
f
als
e
p
o
s
it
iv
e
,
an
d
f
a
ls
e
n
e
g
a
ti
v
e
o
n
t
h
e
d
e
te
cte
d
b
i
n
a
r
y
s
e
g
m
e
n
ta
ti
o
n
m
ask
o
f
a
ta
k
e
n
im
a
g
e
.
H
er
e
we
h
a
v
e
esti
m
at
ed
t
h
e
ev
al
u
a
ti
o
n
m
et
r
i
cs
a
b
o
v
e
f
o
r
e
ac
h
o
f
t
h
e
in
d
i
v
i
d
u
a
l
p
la
n
e
o
f
t
h
e
t
esti
n
g
d
at
ase
t
an
d
we
r
ec
o
r
d
t
h
e
co
m
m
o
n
v
al
u
es
o
f
t
h
e
m
e
n
ti
o
n
e
d
m
et
r
i
cs t
h
r
o
u
g
h
all
i
m
a
g
es
o
f
t
h
e
tes
ti
n
g
s
e
t.
5
.
3
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
PET
o
r
C
T
s
ca
n
s
p
ec
ialis
ts
d
esig
n
ed
p
ix
el
lev
el
s
eg
m
e
n
tatio
n
m
ask
s
f
o
r
i
n
d
iv
id
u
al
3
5
p
atien
ts
to
esti
m
ate
o
u
r
m
o
d
el
wh
ic
h
r
ep
r
esen
t
th
e
test
s
et.
Gr
o
u
n
d
tr
u
t
h
f
r
o
m
th
e
r
ea
l
b
o
u
n
d
in
g
b
o
x
e
s
was
im
p
lem
en
ted
to
tr
ain
th
e
m
o
d
el.
Ho
we
v
er
,
t
h
e
ev
alu
atio
n
m
etr
ics
wer
e
ev
alu
ated
am
o
n
g
th
e
m
o
d
el
d
etec
tio
n
an
d
t
h
e
p
ix
e
l
lev
el
s
eg
m
en
tatio
n
m
ak
es
t
o
a
s
s
u
r
e
th
e
ev
al
u
atio
n
’
s
im
p
o
r
ta
n
t
ef
f
ec
tiv
e
n
ess
.
T
ab
le
3
s
h
o
ws
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
co
m
b
in
atio
n
ap
p
r
o
ac
h
f
o
r
in
d
iv
i
d
u
al
p
lan
e
with
a
9
5
% c
o
n
f
id
en
ce
le
v
el.
Fro
m
T
ab
le
3
,
th
e
p
r
o
p
o
s
ed
R
E
DE
M
-
Net
m
o
d
el
co
n
s
is
ten
tl
y
o
u
tp
er
f
o
r
m
s
alter
n
ativ
e
m
o
d
els
ac
r
o
s
s
f
iv
e
f
o
ld
s
in
f
u
s
in
g
PET
o
r
C
T
p
lan
es.
I
n
f
o
ld
1
,
it
ac
h
iev
e
d
9
8
.
5
5
%
ac
cu
r
ac
y
an
d
7
1
.
2
3
F1
-
s
co
r
e,
s
u
r
p
ass
in
g
m
o
d
ality
-
s
p
ec
if
ic
s
eg
m
en
tatio
n
n
etwo
r
k
(
MSSN
-
Net)
s
ig
n
if
ican
tly
.
I
n
f
o
ld
2
,
it
o
u
tp
er
f
o
r
m
e
d
two
-
s
tag
e
s
eg
m
en
tatio
n
n
etwo
r
k
(
T
DN
N
-
Net
)
with
9
7
.
3
2
%
ac
cu
r
a
cy
an
d
8
0
.
6
9
F1
-
s
co
r
e.
Fo
l
d
3
r
esu
lts
s
h
o
wed
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
o
v
er
c
o
n
tex
t
-
awa
r
e
c
o
n
v
o
lu
tio
n
al
n
etwo
r
k
(
C
AC
-
Net
)
,
wh
ile
f
o
ld
4
also
f
av
o
r
ed
R
E
DE
M
-
Net
o
v
er
T
DNN
-
Net
.
I
n
f
o
ld
5
,
o
u
r
m
o
d
el
r
ea
ch
e
d
9
6
.
6
8
%
ac
c
u
r
ac
y
an
d
8
5
.
6
9
F1
-
s
co
r
e.
Ov
er
all,
R
E
DE
M
-
Net
im
p
r
o
v
es
ac
cu
r
ac
y
b
y
2
0
.
8
1
%,
p
r
ec
is
io
n
b
y
3
2
.
6
4
%,
F1
b
y
1
5
.
3
6
%,
an
d
I
o
U
b
y
2
0
.
3
6
%
o
n
av
er
ag
e.
I
n
F
ig
u
r
e
2
R
E
DE
M
-
Net
m
o
d
el
o
n
ly
u
tili
ze
s
o
n
e
m
o
d
ality
PET
o
r
C
T
,
g
en
er
ally
co
llap
s
e
s
h
o
r
t
o
f
ex
ac
tly
id
en
tify
in
g
th
e
s
ite
o
f
th
e
m
alig
n
an
t
tu
m
o
r
.
T
h
is
ca
n
b
e
clar
if
ied
b
y
th
e
tr
u
th
wh
ich
n
o
t
th
e
f
u
n
ctio
n
a
l
(
PET
)
o
r
th
e
an
at
o
m
ical
(
C
T
)
d
ata
th
at
a
r
e
im
p
o
r
tan
t
f
o
r
d
iag
n
o
s
is
an
d
ca
r
r
ied
i
n
to
ac
co
u
n
t.
PET
a
n
d
C
T
m
o
d
alities
f
u
s
io
n
cr
u
cially
en
h
an
ce
s
th
e
d
iag
n
o
s
tic
ab
il
ities
o
f
th
e
m
o
d
el.
Mo
r
e
o
v
er
,
th
e
p
ix
el
wis
e
s
eg
m
en
tatio
n
m
ask
s
ar
e
en
h
a
n
ce
d
,
an
d
th
ey
ca
n
ex
ac
tly
d
e
f
in
e
th
e
tu
m
o
r
lo
ca
tio
n
an
d
s
i
ze
af
ter
th
e
o
n
lin
e
f
ew
s
h
o
t
r
etr
ain
i
n
g
p
r
o
ce
s
s
will
b
e
im
p
lem
en
ted
.
W
e
u
tili
ze
d
th
e
n
o
n
p
ar
am
etr
ic
W
ilco
x
o
n
s
ig
n
ed
r
an
k
test
th
at
co
n
n
ec
ts
th
e
two
p
air
e
d
g
r
o
u
p
s
to
ex
p
lo
r
e
wh
et
h
er
th
e
F1
an
d
I
o
U
d
ev
elo
p
m
en
t
is
cr
u
cial
as
an
o
u
tco
m
e
o
f
th
e
PET
an
d
C
T
m
o
d
ality
f
u
s
io
n
an
d
t
h
e
p
r
esen
ted
f
ew
s
h
o
ts
r
etr
ain
in
g
a
p
p
r
o
ac
h
.
Ma
jo
r
ly
it
will
b
e
ap
p
lied
in
th
e
lo
ca
tio
n
o
f
t
h
e
p
air
ed
s
tu
d
en
t’
s
test
with
o
u
t
ac
q
u
ir
in
g
th
e
r
ec
o
g
n
ized
d
ata
with
n
o
r
m
al
d
is
tr
ib
u
tio
n
.
W
e
u
tili
ze
d
th
e
W
ilco
x
o
n
test
to
r
elia
b
ly
ac
ce
s
s
th
at
o
u
r
ap
p
r
o
ac
h
h
as
ad
v
a
n
ce
d
F1
an
d
I
o
U
s
c
o
r
e
wh
ile
co
m
p
ar
ed
to
o
th
er
ap
p
r
o
ac
h
es
d
u
e
to
th
e
r
esu
lts
ar
e
co
n
n
ec
ted
a
n
d
d
if
f
er
f
r
o
m
a
n
o
r
m
al
d
is
tr
ib
u
t
io
n
.
W
e
h
a
v
e
ac
h
ie
v
ed
t
h
e
v
a
lu
e
less
th
an
0
.
0
2
f
o
r
th
e
F1
an
d
I
o
U
s
co
r
es
b
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
R
esid
u
a
l e
d
g
e
d
en
s
e
en
h
a
n
ce
d
mo
d
u
le
n
etw
o
r
k:
A
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
…
(
P
r
a
b
a
ka
r
a
n
Ja
ya
r
a
ma
n
)
4039
u
tili
zin
g
th
e
W
ilco
x
o
n
test
.
W
e
will
av
o
id
th
e
h
y
p
o
th
esi
s
with
a
co
n
f
id
en
ce
lev
el
o
f
9
9
.
9
8
%
wh
ile
we
co
m
p
ar
e
th
e
s
u
g
g
ested
R
E
DM
-
Net
m
o
d
el
with
d
if
f
er
en
t a
p
p
r
o
ac
h
es.
Fig
u
r
e
3
p
r
esen
ts
th
e
R
OC
cu
r
v
es
an
d
AUC
v
alu
es
o
f
f
iv
e
m
o
d
els
ev
alu
ate
d
f
o
r
s
em
an
tic
s
eg
m
en
tatio
n
p
er
f
o
r
m
a
n
ce
.
An
AUC
o
f
0
.
5
i
n
d
icate
s
r
an
d
o
m
class
if
icatio
n
,
wh
ile
1
.
0
d
e
n
o
tes
p
er
f
ec
t
ac
cu
r
ac
y
.
Ou
r
p
r
o
p
o
s
ed
R
E
DE
M
-
Net
ac
h
iev
ed
th
e
h
ig
h
est
AUC
o
f
0
.
9
9
8
,
o
u
tp
er
f
o
r
m
in
g
C
AC
-
Ne
t
(
0
.
8
5
6
)
,
T
DNN
-
Net
(
0
.
7
5
6
)
,
MSSN
-
Net
(
0
.
6
2
5
)
,
an
d
MSAM
-
Net
(
0
.
9
2
5
)
,
d
em
o
n
s
tr
atin
g
its
s
u
p
er
io
r
d
iag
n
o
s
tic
ca
p
ab
ilit
y
.
T
ab
le
3
.
Fiv
e
-
f
o
ld
co
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
v
s
ex
is
tin
g
m
o
d
e
l
F
o
l
d
M
o
d
e
l
A
c
c
u
r
a
c
y
±
9
5
%
C
I
P
r
e
c
i
s
i
o
n
±
9
5
%
C
I
R
e
c
a
l
l
±
9
5
%
C
I
F
1
±
9
5
%
C
I
I
O
U
±
9
5
%
C
I
F
o
l
d
1
C
A
C
-
N
e
t
[
8
]
9
6
.
2
1
±
0
.
0
4
6
5
.
3
2
±
2
.
6
2
3
5
.
6
4
±
2
.
1
9
4
3
.
2
1
±
1
.
8
0
3
3
.
2
6
±
0
.
0
6
TD
N
N
-
N
e
t
[
6
]
9
7
.
2
5
±
0
.
0
6
6
9
.
7
5
±
2
.
3
5
5
4
.
3
2
±
2
.
1
5
4
9
.
3
5
±
1
.
6
5
2
9
.
6
5
±
0
.
3
5
M
S
S
N
-
N
e
t
[
1
3
]
9
6
.
5
4
±
0
.
0
7
7
1
.
2
5
±
2
.
4
5
4
9
.
6
8
±
2
.
6
9
5
6
.
3
2
±
2
.
3
5
1
9
.
3
2
±
1
.
2
5
M
S
A
M
-
N
e
t
[
1
4
]
9
4
.
3
2
±
0
.
0
5
7
0
.
6
5
±
2
.
0
5
5
0
.
3
6
±
2
.
2
5
6
0
.
6
5
±
2
.
6
5
2
5
.
6
2
±
2
.
3
6
R
ED
E
M
-
N
ET
9
8
.
5
5
±
0
.
0
5
7
4
.
6
9
±
2
.
9
5
6
5
.
8
7
±
2
.
3
6
7
1
.
2
3
±
3
.
3
0
4
0
.
6
7
±
2
.
9
6
F
o
l
d
2
C
A
C
-
N
e
t
[
8
]
9
5
.
3
6
±
0
.
0
2
7
0
.
2
5
±
2
.
6
5
4
0
.
3
2
±
2
.
6
5
7
0
.
3
2
±
1
.
2
5
1
5
.
3
2
±
0
.
3
5
TD
N
N
-
N
e
t
[
6
]
9
4
.
3
2
±
0
.
3
6
7
2
.
0
5
±
2
.
3
6
4
5
.
0
2
±
3
.
0
6
6
5
.
3
2
±
2
.
3
6
1
9
.
6
5
±
0
.
3
6
M
S
S
N
-
N
e
t
[
1
3
]
9
5
.
6
8
±
0
.
2
8
7
4
.
0
2
±
2
.
9
6
3
9
.
6
5
±
3
.
0
1
4
9
.
2
5
±
3
.
6
5
2
2
.
3
5
±
2
.
3
0
M
S
A
M
-
N
e
t
[
1
4
]
9
4
.
3
5
±
0
.
0
2
7
1
.
3
2
±
1
.
0
2
4
2
.
3
5
±
2
.
6
5
5
5
.
6
2
±
3
.
8
5
3
0
.
8
9
±
2
.
5
8
R
ED
E
M
-
N
ET
9
7
.
3
2
±
0
.
0
2
7
7
.
6
5
±
2
.
6
5
5
5
.
3
2
±
2
.
7
8
8
0
.
6
9
±
2
.
6
5
3
5
.
6
4
±
3
.
3
2
F
o
l
d
3
C
A
C
-
N
e
t
[
8
]
9
6
.
3
2
±
0
.
2
5
6
9
.
3
5
±
2
.
8
7
4
5
.
3
2
±
2
.
1
5
6
9
.
3
2
±
2
.
0
5
2
0
.
3
6
±
0
.
7
5
TD
N
N
-
N
e
t
[
6
]
9
2
.
6
5
±
0
.
3
9
6
0
.
5
5
±
2
.
9
5
3
9
.
6
5
±
2
.
0
3
7
0
.
5
2
±
3
.
8
7
2
5
.
3
7
±
3
.
2
5
M
S
S
N
-
N
e
t
[
1
3
]
9
3
.
4
5
±
0
.
8
7
7
1
.
2
5
±
2
.
6
2
4
2
.
3
5
±
2
.
6
5
7
1
.
3
5
±
1
.
0
2
3
0
.
6
5
±
1
.
0
2
M
S
A
M
-
N
e
t
[
1
4
]
9
4
.
3
5
±
0
.
3
6
7
2
.
3
5
±
2
.
3
9
4
0
.
3
3
±
2
.
9
6
7
2
.
6
5
±
1
.
0
1
2
4
.
9
7
±
2
.
3
6
R
ED
E
M
-
N
ET
9
8
.
6
5
±
0
.
1
5
7
9
.
6
5
±
1
.
0
3
5
0
.
6
5
±
2
.
7
5
8
0
.
6
5
±
2
.
3
6
3
9
.
6
5
±
3
.
6
9
F
o
l
d
4
C
A
C
-
N
e
t
[
8
]
9
3
.
6
5
±
0
.
2
5
7
8
.
6
5
±
2
.
0
3
3
8
.
3
3
±
2
.
8
5
7
2
.
6
5
±
3
.
9
5
1
5
.
3
6
±
2
.
5
8
TD
N
N
-
N
e
t
[
6
]
9
0
.
7
8
±
1
.
2
0
7
0
.
6
5
±
2
.
0
5
3
0
.
0
1
±
3
.
2
5
7
1
.
2
5
±
3
.
7
5
1
9
.
3
7
±
2
.
7
5
M
S
S
N
-
N
e
t
[
1
3
]
9
2
.
6
4
±
1
.
2
6
7
1
.
5
5
±
2
.
8
5
3
2
.
6
5
±
3
.
0
1
4
6
.
3
5
±
3
.
0
5
2
5
.
6
4
±
2
.
1
4
M
S
A
M
-
N
e
t
[
1
4
]
9
0
.
6
5
±
0
.
3
1
7
3
.
6
4
±
2
.
9
6
4
0
.
5
6
±
3
.
0
5
6
5
.
3
9
±
2
.
6
5
1
5
.
3
6
±
2
.
6
5
R
ED
E
M
-
N
ET
9
5
.
6
6
±
0
.
2
0
8
0
.
6
5
±
1
.
2
6
4
5
.
6
4
±
2
.
6
5
7
8
.
3
6
±
2
.
7
5
3
5
.
6
1
±
0
.
5
8
F
o
l
d
5
C
A
C
-
N
e
t
[
8
]
9
2
.
3
5
±
0
.
1
2
7
9
.
3
6
±
2
.
7
5
3
5
.
6
7
±
2
.
6
5
8
0
.
5
6
±
2
.
8
9
2
5
.
6
8
±
0
.
3
1
TD
N
N
-
N
e
t
[
6
]
9
0
.
3
6
±
1
.
3
2
7
5
.
3
6
±
2
.
5
8
4
2
.
9
7
±
2
.
0
5
7
2
.
5
6
±
3
.
0
7
2
0
.
3
1
±
2
.
3
5
M
S
S
N
-
N
e
t
[
1
3
]
9
2
.
8
7
±
2
.
3
6
7
4
.
3
7
±
2
.
4
8
3
9
.
6
7
±
2
.
0
9
8
0
.
2
3
±
2
.
6
5
3
1
.
4
8
±
3
.
2
5
M
S
A
M
-
N
e
t
[
1
4
]
9
4
.
3
6
±
1
.
3
5
7
2
.
8
9
±
2
.
6
0
4
0
.
2
3
±
2
.
3
3
7
5
.
6
2
±
3
.
9
8
3
5
.
6
7
±
3
.
6
9
R
ED
E
M
-
N
ET
9
6
.
6
8
±
2
.
0
1
8
2
.
3
6
±
2
.
7
5
5
0
.
6
7
±
2
.
8
5
8
5
.
6
9
±
3
.
9
9
4
0
.
6
5
±
3
.
9
7
Fig
u
r
e
2
.
Seg
m
e
n
tatio
n
r
esu
lt
co
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
v
s
ex
is
tin
g
wo
r
k
s
Fig
u
r
e
3
.
R
OC
-
AUC illu
s
tr
ati
o
n
o
f
p
r
o
p
o
s
ed
v
s
ex
is
tin
g
wo
r
k
s
5
.
4
.
M
L
ba
s
ed
s
t
a
g
e
cla
s
s
if
ica
t
io
n a
na
ly
s
is
I
n
s
tag
in
g
,
th
e
r
e
wer
e
2
8
p
atien
ts
in
th
e
f
ir
s
t
s
tag
e,
4
5
p
ati
en
ts
in
s
ec
o
n
d
s
tag
e
an
d
9
8
p
atien
ts
in
th
ir
d
s
tag
e.
W
ith
a
co
m
m
o
n
d
ev
elo
p
m
en
t
o
n
4
1
.
0
5
m
o
n
t
h
s
in
th
e
r
a
n
g
e
o
f
1
.
8
-
8
5
.
1
m
o
n
t
h
s
,
th
e
m
e
d
ian
OS
was
1
4
.
3
m
o
n
th
s
in
th
e
r
a
n
g
e
o
f
1
.
2
-
5
1
m
o
n
th
s
.
At
last
,
7
5
p
atien
ts
h
av
e
d
ie
d
(
5
0
%).
C
lass
if
icatio
n
o
u
tco
m
es
o
f
in
d
iv
id
u
al
ML
m
et
h
o
d
s
a
n
d
th
eir
u
n
ity
f
o
ll
o
win
g
t
h
e
tw
o
tech
n
i
q
u
es
ar
e
ex
p
r
ess
ed
in
T
ab
le
4
.
C
o
n
ce
r
n
in
g
th
e
Me
d
ian
OS
en
d
p
o
in
t
in
t
h
e
tr
ain
in
g
s
et,
t
h
e
b
etter
m
o
d
el
co
n
s
tr
u
cte
d
b
y
DT
with
2
0
f
ea
tu
r
es
th
at
h
av
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
0
3
2
-
4
0
4
2
4040
ac
h
iev
ed
a
n
ac
c
u
r
ac
y
8
5
%
[
8
0
%
C
I
5
5
-
7
1
]
i
n
th
e
tr
ain
in
g
s
et.
I
n
th
e
test
in
g
s
et,
DT
h
as
ac
h
iev
ed
6
7
%
[
9
5
% C
I
5
2
-
6
1
]
ac
c
u
r
ac
y
.
T
ab
le
4
.
Stag
e
class
if
icatio
n
c
o
m
p
ar
is
o
n
a
m
o
n
g
ML
m
o
d
els
A
c
c
u
r
a
c
y
(
M
e
d
i
a
n
O
S
%)
B
a
l
a
n
c
e
d
a
c
c
u
r
a
c
y
(
O
S
<
6
mo
n
t
h
s
%)
C
l
a
s
si
f
i
c
a
t
i
o
n
Tr
a
i
n
i
n
g
Te
st
i
n
g
N
o
o
f
p
a
t
i
e
n
t
s
p
r
o
p
e
r
l
y
c
l
a
ssi
f
i
e
d
Tr
a
i
n
i
n
g
Te
st
i
n
g
N
o
o
f
p
a
t
i
e
n
t
s
p
r
o
p
e
r
l
y
c
l
a
ssi
f
i
e
d
D
T
85
[
8
0
C
I
5
5
-
71]
67
[
9
5
C
I
5
2
-
61]
30
59
[
9
5
C
I
6
5
-
78]
51
[
9
5
C
I
5
0
-
55]
26
R
F
88
[
9
5
C
I
6
5
–
8
1
]
66
[
9
5
C
I
6
0
–
7
0
]
33
85
[
9
5
C
I
5
2
-
61]
59
[
9
5
C
I
6
0
-
70]
36
X
G
B
o
o
st
90
[
9
5
C
I
9
0
–
9
5
]
65
[
9
5
C
I
5
0
–
6
5
]
32
72
[
9
5
C
I
6
1
-
75]
55
[
9
5
C
I
7
5
-
80]
40
LR
87
[
9
5
C
I
6
2
-
75]
62
[
9
5
C
I
5
0
–
7
0
]
35
80
[
9
5
C
I
5
1
-
71]
65
[
9
5
C
I
5
6
-
71]
42
MC
-
S
V
M
95
[
9
5
C
I
9
5
-
95]
75
[
9
5
C
I
6
5
-
76]
37
92
[
9
5
C
I
7
5
-
86]
90
[
9
5
C
I
6
5
-
76]
45
I
n
ad
d
itio
n
,
th
e
n
e
x
t
m
o
d
el
R
F
h
as
co
m
b
in
ed
with
2
5
f
e
atu
r
es
an
d
ac
h
iev
e
d
an
ac
c
u
r
ac
y
o
f
8
8
%
[
9
5
% C
I
6
5
–
8
1
]
in
tr
ain
in
g
s
et.
I
n
th
e
test
in
g
s
et,
R
F a
ch
iev
e
d
6
6
% [
9
5
% C
I
6
0
–
7
0
]
ac
cu
r
a
cy
.
Mo
r
eo
v
er
,
XG
B
o
o
s
t
co
m
b
in
ed
with
3
0
f
ea
t
u
r
es
an
d
attain
ed
a
n
ac
cu
r
a
cy
o
f
9
0
%
[
9
5
%
C
I
9
0
–
9
5
]
in
th
e
tr
ain
in
g
s
et.
I
n
th
e
test
in
g
s
et,
XG
b
o
o
s
t
ac
h
iev
ed
6
5
%
[
9
5
%
C
I
5
0
–
6
5
]
o
f
ac
cu
r
ac
y
.
L
R
co
m
b
in
ed
wit
h
3
2
f
ea
tu
r
es
an
d
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
8
7
%
[
9
5
% C
I
6
2
-
7
5
]
i
n
tr
ain
in
g
s
et.
I
n
th
e
test
in
g
s
et,
L
R
ac
h
iev
ed
6
2
%
[
9
5
%
C
I
5
0
–
7
0
]
o
f
ac
cu
r
ac
y
.
At
last
,
o
u
r
s
u
g
g
ested
MC
-
SVM
h
as
co
m
b
in
ed
with
4
0
f
ea
tu
r
es
an
d
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
5
%
[
9
5
%
C
I
9
5
-
9
5
]
in
tr
ain
in
g
test
.
I
n
th
e
test
in
g
s
et,
M
C
-
SVM
h
as
a
t
tain
ed
7
5
%
[
9
5
%
C
I
6
5
-
7
6
]
o
f
ac
cu
r
ac
y
.
T
h
is
m
ea
n
s
th
at
th
e
in
d
iv
id
u
al
alg
o
r
ith
m
s
s
ep
ar
ately
p
r
o
p
er
l
y
d
etec
ted
th
e
p
atien
t’
s
s
u
r
v
iv
al
ab
o
v
e
o
r
b
elo
w
v
alu
e
f
o
r
th
e
m
en
tio
n
e
d
alg
o
r
ith
m
s
.
B
y
co
m
p
a
r
in
g
th
ese
ap
p
r
o
ac
h
es,
o
u
r
s
u
g
g
este
d
ap
p
r
o
ac
h
h
as
r
ea
ch
e
d
h
ig
h
er
ac
cu
r
ac
y
i
n
th
e
tr
ain
in
g
s
et
an
d
th
e
test
in
g
s
e
t.
C
o
n
ce
r
n
in
g
th
e
s
ec
o
n
d
en
d
p
o
in
t
o
n
OS
b
elo
w
6
m
o
n
th
s
wh
ile
co
m
p
ar
ed
to
o
th
er
a
p
p
r
o
ac
h
es
o
u
r
s
u
g
g
est
ed
m
o
d
el
MC
-
SVM
h
as
ac
h
iev
ed
h
ig
h
er
ac
cu
r
ac
y
9
2
%
[
9
5
%
C
I
7
5
-
8
6
]
i
n
tr
ain
in
g
s
et.
I
n
t
h
e
test
in
g
s
et,
MC
-
SVM
h
as a
ttain
ed
9
0
% [
9
5
% C
I
6
5
-
7
6
]
o
f
ac
cu
r
ac
y
.
6.
CO
NCLU
SI
O
N
W
e
d
ev
elo
p
a
R
E
DE
M
-
NE
T
tech
n
iq
u
e
f
o
r
ac
cu
r
ate
s
eg
m
en
t
an
d
class
if
y
lu
n
g
tu
m
o
r
s
tag
e
b
y
p
r
o
ce
s
s
in
g
PET
an
d
C
T
im
ag
es.
I
n
itially
,
DR
C
N
to
co
llect
in
p
u
t
d
ata
an
d
ex
tr
ac
t
h
i
g
h
-
d
im
en
s
io
n
al
f
ea
tu
r
es
s
im
u
ltan
eo
u
s
ly
.
T
h
en
th
e
e
x
tr
ac
ted
f
ea
tu
r
es
o
f
b
o
th
im
a
g
es
ar
e
th
en
p
ass
ed
in
to
UNe
t++
to
ac
q
u
ir
e
m
u
lti
-
lev
el
d
ec
o
d
e
d
f
ea
tu
r
es.
T
h
e
d
ec
o
d
e
d
f
ea
tu
r
e
is
f
u
r
th
er
p
r
o
ce
s
s
ed
th
r
o
u
g
h
two
s
p
ec
ia
lized
m
o
d
u
les,
th
e
P
EL
M
an
d
E
2
L
M
in
o
r
d
er
to
cr
ea
te
ac
cu
r
ate
tu
m
o
r
s
eg
m
e
n
tatio
n
.
Fin
ally
,
th
e
o
u
tp
u
ts
o
f
th
ese
m
o
d
u
les
ar
e
m
er
g
ed
t
o
p
r
o
d
u
ce
a
p
r
ec
is
e
s
eg
m
en
tatio
n
.
T
h
en
th
e
s
eg
m
e
n
ted
tu
m
o
r
is
class
if
ied
in
to
s
tag
es
b
y
u
tili
zin
g
a
MC
-
S
VM
to
id
e
n
tifie
s
p
r
im
ar
y
tu
m
o
r
,
r
eg
io
n
ly
m
p
h
n
o
d
e
an
d
d
is
tan
t m
etastas
is
to
clas
s
if
y
lu
n
g
tu
m
o
r
s
tag
es.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
t
h
o
r
s
ac
k
n
o
wled
g
e
SR
M
I
n
s
titu
te
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
(
SR
MI
ST)
f
o
r
p
r
o
v
id
in
g
th
e
n
ec
ess
ar
y
f
ac
ilit
ies to
ca
r
r
y
o
u
t th
is
r
esear
ch
an
d
th
a
n
k
th
e
r
e
v
iewe
r
s
f
o
r
th
eir
c
o
n
s
tr
u
ctiv
e
s
u
g
g
esti
o
n
s
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
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
Pra
b
ak
ar
an
J
ay
a
r
am
an
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Pan
d
iar
aj
Selv
ar
aj
✓
✓
✓
✓
✓
✓
✓
✓
Ash
win
i E
lan
g
o
✓
✓
✓
✓
✓
✓
✓
✓
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
R
esid
u
a
l e
d
g
e
d
en
s
e
en
h
a
n
ce
d
mo
d
u
le
n
etw
o
r
k:
A
d
ee
p
lea
r
n
in
g
a
p
p
r
o
a
c
h
…
(
P
r
a
b
a
ka
r
a
n
Ja
ya
r
a
ma
n
)
4041
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
th
at
th
e
y
h
av
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
d
ata,
wh
ich
c
o
n
tain
in
f
o
r
m
atio
n
th
at
co
u
l
d
co
m
p
r
o
m
is
e
th
e
p
r
iv
ac
y
o
f
r
esear
c
h
p
ar
tic
ip
an
ts
,
ar
e
n
o
t p
u
b
licly
av
ailab
le
d
u
e
to
c
er
tain
r
estrictio
n
s
.
R
E
FE
R
E
N
C
E
S
[
1
]
R
.
F
u
j
i
k
a
w
a
e
t
a
l
.
,
“
C
l
i
n
i
c
o
p
a
t
h
o
l
o
g
i
c
a
n
d
g
e
n
o
t
y
p
i
c
f
e
a
t
u
r
e
s
o
f
l
u
n
g
a
d
e
n
o
c
a
r
c
i
n
o
m
a
c
h
a
r
a
c
t
e
r
i
z
e
d
b
y
t
h
e
i
n
t
e
r
n
a
t
i
o
n
a
l
a
sso
c
i
a
t
i
o
n
f
o
r
t
h
e
st
u
d
y
o
f
l
u
n
g
c
a
n
c
e
r
g
r
a
d
i
n
g
sy
s
t
e
m
,
”
J
o
u
r
n
a
l
o
f
T
h
o
r
a
c
i
c
O
n
c
o
l
o
g
y
,
v
o
l
.
1
7
,
n
o
.
5
,
p
p
.
7
0
0
–
7
0
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
t
h
o
.
2
0
2
2
.
0
2
.
0
0
5
.
[
2
]
R
.
R
a
m
i
-
P
o
r
t
a
,
“
F
u
t
u
r
e
p
e
r
s
p
e
c
t
i
v
e
s
o
n
t
h
e
t
n
m
s
t
a
g
i
n
g
f
o
r
l
u
n
g
c
a
n
c
e
r
,
”
C
a
n
c
e
rs
,
v
o
l
.
1
3
,
n
o
.
8
,
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
c
a
n
c
e
r
s
1
3
0
8
1
9
4
0
.
[
3
]
R
.
U
.
O
saro
g
i
a
g
b
o
n
e
t
a
l
.
,
“
T
h
e
i
n
t
e
r
n
a
t
i
o
n
a
l
a
ss
o
c
i
a
t
i
o
n
f
o
r
t
h
e
st
u
d
y
o
f
l
u
n
g
c
a
n
c
e
r
l
u
n
g
c
a
n
c
e
r
s
t
a
g
i
n
g
p
r
o
j
e
c
t
:
o
v
e
r
v
i
e
w
o
f
c
h
a
l
l
e
n
g
e
s
a
n
d
o
p
p
o
r
t
u
n
i
t
i
e
s
i
n
r
e
v
i
si
n
g
t
h
e
n
o
d
a
l
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
l
u
n
g
c
a
n
c
e
r
,
”
J
o
u
rn
a
l
o
f
T
h
o
r
a
c
i
c
O
n
c
o
l
o
g
y
,
v
o
l
.
1
8
,
n
o
.
4
,
p
p
.
4
1
0
–
4
1
8
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
t
h
o
.
2
0
2
2
.
1
2
.
0
0
9
.
[
4
]
Y
.
O
h
n
o
e
t
a
l
.
,
“
S
ma
l
l
c
e
l
l
l
u
n
g
c
a
n
c
e
r
st
a
g
i
n
g
:
p
r
o
s
p
e
c
t
i
v
e
c
o
m
p
a
r
i
so
n
o
f
c
o
n
v
e
n
t
i
o
n
a
l
s
t
a
g
i
n
g
t
e
s
t
s,
F
D
G
P
ET
/
C
T
,
w
h
o
l
e
-
b
o
d
y
M
R
I
,
a
n
d
c
o
r
e
g
i
s
t
e
r
e
d
F
D
G
P
E
T/
M
R
I
,
”
Am
e
ri
c
a
n
J
o
u
r
n
a
l
o
f
Ro
e
n
t
g
e
n
o
l
o
g
y
,
v
o
l
.
2
1
8
,
n
o
.
5
,
p
p
.
8
9
9
–
9
0
8
,
2
0
2
2
,
d
o
i
:
1
0
.
2
2
1
4
/
A
JR
.
2
1
.
2
6
8
6
8
.
[
5
]
N
.
M
.
B
a
t
o
u
t
y
e
t
a
l
.
,
“
S
t
a
t
e
o
f
t
h
e
a
r
t
:
l
u
n
g
c
a
n
c
e
r
s
t
a
g
i
n
g
u
s
i
n
g
u
p
d
a
t
e
d
i
ma
g
i
n
g
mo
d
a
l
i
t
i
e
s,”
Bi
o
e
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
9
,
n
o
.
1
0
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
b
i
o
e
n
g
i
n
e
e
r
i
n
g
9
1
0
0
4
9
3
.
[
6
]
J.
P
a
r
k
e
t
a
l
.
,
“
A
u
t
o
mat
i
c
l
u
n
g
c
a
n
c
e
r
seg
m
e
n
t
a
t
i
o
n
i
n
[
1
8
F
]
F
D
G
P
ET/
C
T
u
s
i
n
g
a
t
w
o
-
st
a
g
e
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
N
u
c
l
e
a
r
Me
d
i
c
i
n
e
a
n
d
M
o
l
e
c
u
l
a
r Im
a
g
i
n
g
,
v
o
l
.
5
7
,
n
o
.
2
,
p
p
.
8
6
–
9
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
3
1
3
9
-
022
-
0
0
7
4
5
-
7.
[
7
]
H
.
S
h
e
n
e
t
a
l
.
,
“
A
s
u
b
r
e
g
i
o
n
-
b
a
s
e
d
p
o
s
i
t
r
o
n
e
m
i
ssi
o
n
t
o
m
o
g
r
a
p
h
y
/
c
o
mp
u
t
e
d
t
o
mo
g
r
a
p
h
y
(
P
ET/
C
T)
r
a
d
i
o
m
i
c
s
m
o
d
e
l
f
o
r
t
h
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
n
o
n
-
sma
l
l
c
e
l
l
l
u
n
g
c
a
n
c
e
r
h
i
st
o
p
a
t
h
o
l
o
g
i
c
a
l
su
b
t
y
p
e
s,
”
Q
u
a
n
t
i
t
a
t
i
v
e
I
m
a
g
i
n
g
i
n
M
e
d
i
c
i
n
e
a
n
d
S
u
r
g
e
ry
,
v
o
l
.
1
1
,
n
o
.
7
,
p
p
.
2
9
1
8
–
2
9
3
2
,
2
0
2
1
,
d
o
i
:
1
0
.
2
1
0
3
7
/
q
i
ms
-
20
-
1
1
8
2
.
[
8
]
D
.
El
H
a
m
d
i
,
I
.
El
o
u
e
d
i
,
a
n
d
I
.
S
l
i
m,
“
C
o
mp
u
t
e
r
-
a
i
d
e
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
c
e
l
l
l
u
n
g
c
a
n
c
e
r
v
i
a
P
ET
/
C
T
i
m
a
g
e
s
u
si
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
m
a
g
e
a
n
d
G
r
a
p
h
i
c
s
,
v
o
l
.
2
4
,
n
o
.
4
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
4
2
/
S
0
2
1
9
4
6
7
8
2
4
5
0
0
4
0
2
.
[
9
]
G
.
K
a
si
n
a
t
h
a
n
a
n
d
S
.
Ja
y
a
k
u
m
a
r
,
“
C
l
o
u
d
-
b
a
s
e
d
l
u
n
g
t
u
m
o
r
d
e
t
e
c
t
i
o
n
a
n
d
st
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s,
”
Bi
o
Me
d
Re
se
a
r
c
h
I
n
t
e
r
n
a
t
i
o
n
a
l
,
v
o
l
.
2
0
2
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
4
1
8
5
8
3
5
.
[
1
0
]
K
.
B
a
r
b
o
u
c
h
i
,
D
.
El
H
a
md
i
,
I
.
El
o
u
e
d
i
,
T.
B
e
n
A
ï
c
h
a
,
A
.
K
.
Ec
h
i
,
a
n
d
I
.
S
l
i
m,
“
A
t
r
a
n
sf
o
r
mer
-
b
a
se
d
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
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
o
f
l
u
n
g
c
a
n
c
e
r
v
i
a
P
ET
/
C
T
i
m
a
g
e
s,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
m
a
g
i
n
g
S
y
s
t
e
m
s
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
3
3
,
n
o
.
4
,
p
p
.
1
3
8
3
–
1
3
9
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
2
/
i
ma.
2
2
8
5
8
.
[
1
1
]
D
.
Q
.
W
a
n
g
e
t
a
l
.
,
“
A
ss
e
ssi
n
g
d
y
n
a
mi
c
me
t
a
b
o
l
i
c
h
e
t
e
r
o
g
e
n
e
i
t
y
i
n
n
o
n
-
smal
l
c
e
l
l
l
u
n
g
c
a
n
c
e
r
p
a
t
i
e
n
t
s
v
i
a
u
l
t
r
a
-
h
i
g
h
se
n
si
t
i
v
i
t
y
t
o
t
a
l
-
b
o
d
y
[
1
8
F
]
F
D
G
P
ET/
C
T
i
m
a
g
i
n
g
:
q
u
a
n
t
i
t
a
t
i
v
e
a
n
a
l
y
s
i
s
o
f
[
1
8
F
]
F
D
G
u
p
t
a
k
e
i
n
p
r
i
mar
y
t
u
mo
r
s
a
n
d
met
a
s
t
a
t
i
c
l
y
mp
h
n
o
d
e
s
,
”
E
u
r
o
p
e
a
n
J
o
u
rn
a
l
o
f
N
u
c
l
e
a
r
M
e
d
i
c
i
n
e
a
n
d
M
o
l
e
c
u
l
a
r
I
m
a
g
i
n
g
,
v
o
l
.
4
9
,
n
o
.
1
3
,
p
p
.
4
6
9
2
–
4
7
0
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
2
5
9
-
0
2
2
-
0
5
9
0
4
-
8.
[
1
2
]
J.
D
.
R
o
se
,
K
.
J
a
sp
i
n
,
a
n
d
K
.
V
i
j
a
y
a
k
u
m
a
r
,
“
Lu
n
g
c
a
n
c
e
r
d
i
a
g
n
o
si
s
b
a
s
e
d
o
n
i
ma
g
e
f
u
si
o
n
a
n
d
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
C
T
a
n
d
P
E
T
i
ma
g
e
,
”
i
n
S
i
g
n
a
l
a
n
d
I
m
a
g
e
Pro
c
e
ss
i
n
g
T
e
c
h
n
i
q
u
e
s
f
o
r
t
h
e
D
e
v
e
l
o
p
m
e
n
t
o
f
I
n
t
e
l
l
i
g
e
n
t
H
e
a
l
t
h
c
a
re
S
y
s
t
e
m
s
,
S
i
n
g
a
p
o
r
e
:
S
p
r
i
n
g
e
r
,
2
0
2
1
,
p
p
.
6
7
–
86
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
981
-
15
-
6
1
4
1
-
2
_
4
.
[
1
3
]
D
.
X
i
a
n
g
,
B
.
Zh
a
n
g
,
Y
.
L
u
,
a
n
d
S
.
D
e
n
g
,
“
M
o
d
a
l
i
t
y
-
sp
e
c
i
f
i
c
s
e
g
me
n
t
a
t
i
o
n
n
e
t
w
o
r
k
f
o
r
l
u
n
g
t
u
m
o
r
seg
me
n
t
a
t
i
o
n
i
n
P
ET
-
C
T
i
ma
g
e
s,
”
I
EEE
J
o
u
rn
a
l
o
f
B
i
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
7
,
n
o
.
3
,
p
p
.
1
2
3
7
–
1
2
4
8
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
JB
H
I
.
2
0
2
2
.
3
1
8
6
2
7
5
.
[
1
4
]
X
.
F
u
,
L.
B
i
,
A
.
K
u
mar
,
M
.
F
u
l
h
a
m,
a
n
d
J.
K
i
m,
“
M
u
l
t
i
mo
d
a
l
sp
a
t
i
a
l
a
t
t
e
n
t
i
o
n
m
o
d
u
l
e
f
o
r
t
a
r
g
e
t
i
n
g
m
u
l
t
i
m
o
d
a
l
P
ET
-
C
T
l
u
n
g
t
u
m
o
r
se
g
m
e
n
t
a
t
i
o
n
,
”
I
E
EE
J
o
u
rn
a
l
o
f
B
i
o
m
e
d
i
c
a
l
a
n
d
H
e
a
l
t
h
I
n
f
o
r
m
a
t
i
c
s
,
v
o
l
.
2
5
,
n
o
.
9
,
p
p
.
3
5
0
7
–
3
5
1
6
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
JB
H
I
.
2
0
2
1
.
3
0
5
9
4
5
3
.
[
1
5
]
Y
.
X
i
e
e
t
a
l
.
,
“
A
P
ET/
C
T
n
o
m
o
g
r
a
m
i
n
c
o
r
p
o
r
a
t
i
n
g
S
U
V
ma
x
a
n
d
C
T
r
a
d
i
o
mi
c
s
f
o
r
p
r
e
o
p
e
r
a
t
i
v
e
n
o
d
a
l
st
a
g
i
n
g
i
n
n
o
n
-
sma
l
l
c
e
l
l
l
u
n
g
c
a
n
c
e
r
,
”
E
u
r
o
p
e
a
n
Ra
d
i
o
l
o
g
y
,
v
o
l
.
3
1
,
n
o
.
8
,
p
p
.
6
0
3
0
–
6
0
3
8
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
3
3
0
-
0
2
0
-
0
7
6
2
4
-
9.
[
1
6
]
X
.
X
i
a
a
n
d
R
.
Z
h
a
n
g
,
“
A
n
o
v
e
l
l
u
n
g
n
o
d
u
l
e
a
c
c
u
r
a
t
e
se
g
m
e
n
t
a
t
i
o
n
o
f
P
E
T
-
C
T
i
m
a
g
e
s b
a
s
e
d
o
n
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
a
n
d
g
r
a
p
h
mo
d
e
l
,
”
I
EE
E
A
c
c
e
s
s
,
v
o
l
.
1
1
,
p
p
.
3
4
0
1
5
–
3
4
0
3
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
2
6
2
7
2
9
.
[
1
7
]
N
.
N
a
w
r
e
e
n
,
U
.
H
a
n
y
,
a
n
d
T
.
I
sl
a
m,
“
L
u
n
g
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
s
si
f
i
c
a
t
i
o
n
u
si
n
g
C
T
sc
a
n
i
m
a
g
e
p
r
o
c
e
ssi
n
g
,
”
i
n
2
0
2
1
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
A
u
t
o
m
a
t
i
o
n
,
C
o
n
t
ro
l
a
n
d
Me
c
h
a
t
ro
n
i
c
s
f
o
r
I
n
d
u
st
r
y
4
.
0
(
A
C
MI)
,
2
0
2
1
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
M
I
5
3
8
7
8
.
2
0
2
1
.
9
5
2
8
2
9
7
.
[
1
8
]
A
.
R
e
h
ma
n
,
M
.
K
a
s
h
i
f
,
I
.
A
b
u
n
a
d
i
,
a
n
d
N
.
A
y
e
s
h
a
,
“
L
u
n
g
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
s
si
f
i
c
a
t
i
o
n
f
r
o
m
c
h
e
s
t
C
T
s
c
a
n
s u
si
n
g
mac
h
i
n
e
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s,
”
i
n
2
0
2
1
1
st
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
A
rt
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
An
a
l
y
t
i
c
s
,
C
AI
D
A
2
0
2
1
,
2
0
2
1
,
p
p
.
1
0
1
–
104
,
d
o
i
:
1
0
.
1
1
0
9
/
C
A
I
D
A
5
1
9
4
1
.
2
0
2
1
.
9
4
2
5
2
6
9
.
[
1
9
]
P
.
Y
a
d
a
v
,
N
.
M
e
n
o
n
,
V
.
R
a
v
i
,
a
n
d
S
.
V
i
sh
v
a
n
a
t
h
a
n
,
“
L
u
n
g
-
G
A
N
s:
U
n
su
p
e
r
v
i
s
e
d
r
e
p
r
e
se
n
t
a
t
i
o
n
l
e
a
r
n
i
n
g
f
o
r
l
u
n
g
d
i
se
a
s
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
c
h
e
s
t
C
T
a
n
d
X
-
r
a
y
i
ma
g
e
s,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s o
n
En
g
i
n
e
e
ri
n
g
M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
7
0
,
n
o
.
8
,
p
p
.
2
7
7
4
–
2
7
8
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TE
M
.
2
0
2
1
.
3
1
0
3
3
3
4
.
[
2
0
]
R
.
R
a
z
a
e
t
a
l
.
,
“
L
u
n
g
-
Ef
f
N
e
t
:
L
u
n
g
c
a
n
c
e
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
Ef
f
i
c
i
e
n
t
N
e
t
f
r
o
m
C
T
-
sca
n
i
ma
g
e
s
,
”
E
n
g
i
n
e
e
ri
n
g
Ap
p
l
i
c
a
t
i
o
n
s
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
2
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
a
p
p
a
i
.
2
0
2
3
.
1
0
6
9
0
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.