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.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
,
p
p
.
2471
~
2
4
8
3
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
4
7
1
-
2
4
8
3
2471
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Co
mputed
tom
o
g
ra
phy
ima
g
ing
radio
mics
:
a
nov
el a
ppro
a
ch t
o
ea
rly
-
sta
g
e non
-
s
ma
ll cell
lung
ca
ncer pre
diction
Ra
v
it
ej
a
B
a
le
k
a
i
1
,
M
a
llik
a
rj
un
S.
H
o
li
2
1
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
G
M
I
n
st
i
t
u
t
e
o
f
T
e
c
h
n
o
l
o
g
y
,
V
i
sv
e
s
v
a
r
a
y
a
Te
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
si
t
y
,
B
e
l
a
g
a
v
i
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
I
n
st
r
u
men
t
a
t
i
o
n
E
n
g
i
n
e
e
r
i
n
g
,
U
n
i
v
e
r
si
t
y
B
.
D
.
T
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
,
V
i
sv
e
sv
a
r
a
y
a
T
e
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
s
i
t
y
,
B
e
l
a
g
a
v
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
No
v
1
9
,
2
0
2
4
R
ev
is
ed
Feb
1
1
,
2
0
2
6
Acc
ep
ted
Ma
y
1
1
,
2
0
2
6
Ra
d
io
m
ics
sh
o
ws
p
r
o
m
ise
a
s
n
o
n
-
i
n
v
a
siv
e
m
e
th
o
d
f
o
r
e
n
h
a
n
c
i
n
g
c
li
n
ica
l
sta
g
in
g
o
f
n
o
n
-
sm
a
ll
c
e
ll
l
u
n
g
c
a
n
c
e
r
(NSCL
C)
b
y
u
sin
g
q
u
a
n
ti
tati
v
e
in
fo
rm
a
ti
o
n
fro
m
c
o
m
p
u
ted
to
m
o
g
ra
p
h
y
(CT)
sc
a
n
s.
Th
is
stu
d
y
p
re
se
n
ts
ra
d
io
m
ics
-
b
a
se
d
m
a
c
h
in
e
lea
rn
i
n
g
(M
L)
a
p
p
r
o
a
c
h
fo
r
sta
g
in
g
NSCL
C
p
a
ti
e
n
ts
in
t
o
c
li
n
ica
l
sta
g
e
s
I,
II
,
a
n
d
III
b
a
se
d
o
n
s
h
a
p
e
,
in
te
n
sity
,
a
n
d
tex
tu
re
fe
a
tu
re
s.
CT
ima
g
e
s
o
f
3
6
9
NSCL
C
p
a
ti
e
n
ts
a
re
c
o
ll
e
c
ted
fro
m
th
e
c
a
n
c
e
r
ima
g
in
g
a
rc
h
iv
e
(TCIA
),
a
n
d
e
x
t
ra
c
ted
1
0
7
ra
d
i
o
m
ic
fe
a
tu
re
s
fo
ll
o
wi
n
g
ima
g
e
b
i
o
m
a
rk
e
r
sta
n
d
a
rd
iza
ti
o
n
in
it
iativ
e
(IBS
I)
p
r
o
t
o
c
o
l.
T
h
e
a
n
a
ly
sis
o
f
th
e
so
u
rc
e
s
o
f
v
a
riab
il
it
y
d
u
e
t
o
d
iffere
n
t
ima
g
i
n
g
p
ro
t
o
c
o
ls,
u
sin
g
p
rin
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
ly
sis
(P
CA)
a
n
d
t
-
d
istri
b
u
ted
sto
c
h
a
stic
n
e
ig
h
b
o
r
e
m
b
e
d
d
i
n
g
(t
-
S
NE)
,
s
h
o
we
d
th
a
t
t
h
e
se
e
ffe
c
ts
we
re
re
so
lv
e
d
th
ro
u
g
h
Co
m
Ba
t
h
a
rm
o
n
iza
ti
o
n
.
R
e
c
u
rsiv
e
fe
a
tu
re
e
li
m
in
a
ti
o
n
(
RF
E)
a
n
d
lea
st
a
b
so
lu
te
sh
rin
k
a
g
e
a
n
d
se
lec
ti
o
n
o
p
e
ra
to
r
(LA
S
S
O)
a
re
u
se
d
fo
r
fe
a
tu
re
se
lec
ti
o
n
.
F
iv
e
M
L
a
lg
o
rit
h
m
s
:
lo
g
isti
c
re
g
re
ss
io
n
(LR)
,
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
(S
VM)
,
d
e
c
isio
n
tree
(DT)
,
ra
n
d
o
m
f
o
re
st
(R
F
),
a
n
d
e
x
trem
e
g
ra
d
ien
t
b
o
o
sti
n
g
(
XG
Bo
o
st
)
we
re
u
se
d
,
with
a
n
8
0
:
2
0
train
-
tes
t
sp
li
t
a
n
d
10
-
f
o
ld
c
ro
ss
-
v
a
li
d
a
ti
o
n
.
T
h
e
c
las
sifier
is
a
ss
e
ss
e
d
u
sin
g
a
c
c
u
ra
c
y
,
se
n
siti
v
it
y
,
s
p
e
c
ifi
c
it
y
,
F
1
-
sc
o
re
,
a
n
d
a
re
a
u
n
d
e
r
th
e
re
c
e
iv
e
r
o
p
e
ra
ti
n
g
c
h
a
ra
c
teristic
(AU
ROC)
c
u
rv
e
.
Th
e
RF
E
a
n
d
RF
c
las
sifier
c
o
m
b
in
a
ti
o
n
p
e
rfo
rm
e
d
th
e
b
e
st
with
AU
RO
C
o
f
0
.
9
3
0
7
a
n
d
a
c
c
u
ra
c
y
o
f
0
.
8
1
1
4
.
T
h
is
stu
d
y
i
ll
u
stra
tes
th
e
u
se
o
f
ra
d
i
o
m
ics
m
o
d
e
ls i
n
n
o
n
-
in
v
a
siv
e
c
las
sifica
ti
o
n
o
f
NSCL
C
sta
g
e
s a
n
d
it
i
s r
o
le i
n
c
li
n
ica
l
d
e
c
isio
n
m
a
k
in
g
.
K
ey
w
o
r
d
s
:
C
lin
ical
s
tag
e
C
o
m
p
u
ted
to
m
o
g
r
a
p
h
y
Featu
r
e
s
elec
tio
n
Ma
ch
in
e
lear
n
in
g
No
n
-
s
m
all
ce
ll lu
n
g
ca
n
ce
r
R
ad
io
m
ics
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
:
R
av
iteja
B
alek
ai
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
,
GM
I
n
s
titu
te
o
f
T
ec
h
n
o
lo
g
y
Vis
v
esv
ar
ay
a
T
ec
h
n
o
l
o
g
ical
Un
iv
er
s
ity
B
elag
av
i,
I
n
d
ia
E
m
ail: r
av
itejj1
0
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
L
u
n
g
ca
n
ce
r
is
th
e
m
o
s
t
co
m
m
o
n
ca
u
s
e
o
f
ca
n
ce
r
-
r
elate
d
m
o
r
tality
in
th
e
w
o
r
ld
,
an
d
in
2
0
2
2
alo
n
e
it
was
esti
m
ated
to
ca
u
s
e
2
.
4
8
m
illi
o
n
n
ew
ca
s
es
an
d
1
.
8
1
m
illi
o
n
d
ea
th
s
[
1
]
.
T
h
e
m
o
s
t
co
m
m
o
n
s
u
b
ty
p
e
f
o
r
m
is
non
-
s
m
all
ce
ll lu
n
g
ca
n
ce
r
(
NSC
L
C
)
,
wh
ich
co
n
s
titu
tes ap
p
r
o
x
im
ately
8
5
%
o
f
th
e
ca
s
es [
2
]
,
an
d
it in
clu
d
es
ad
en
o
ca
r
cin
o
m
a
(
ADC),
s
q
u
a
m
o
u
s
ce
ll
ca
r
cin
o
m
a
(
SC
C
)
,
an
d
lar
g
e
ce
ll
ca
r
cin
o
m
a
(
L
C
C
)
,
ea
ch
with
it
i
s
o
wn
b
io
l
o
g
ical
b
e
h
av
io
r
an
d
tr
ea
tm
en
t
r
esp
o
n
s
e.
Alth
o
u
g
h
d
iag
n
o
s
is
an
d
t
r
ea
tm
en
t
h
av
e
ad
v
a
n
ce
d
,
th
e
m
o
r
tality
r
ate
o
f
lu
n
g
ca
n
ce
r
r
em
ain
s
h
ig
h
,
an
d
th
e
5
-
y
ea
r
s
u
r
v
iv
al
o
f
th
e
NSC
L
C
p
atien
ts
is
ab
o
u
t 2
6
.
9
% [
3
]
.
E
ar
ly
d
ia
g
n
o
s
is
o
f
NSC
L
C
is
ess
en
tial
in
f
ac
ilit
atin
g
c
u
r
ativ
e
th
er
ap
y
,
wh
ich
ca
n
s
ig
n
if
ican
tly
e
n
h
an
ce
p
atien
t o
u
tco
m
es.
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.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
4
7
1
-
2
4
8
3
2472
T
h
e
s
tag
in
g
o
f
NSC
L
C
i
s
b
as
ed
o
n
th
e
T
NM
class
if
icatio
n
s
y
s
tem
,
wh
ich
ass
e
s
s
e
s
th
e
s
ize
o
f
th
e
tu
m
o
r
(
T
)
,
ly
m
p
h
n
o
d
e
(
N)
,
an
d
m
etastas
is
(
M)
.
T
h
e
s
tag
es
ar
e
I
(
lo
ca
lized
tu
m
o
r
)
,
I
I
(
s
p
r
ea
d
to
ly
m
p
h
n
o
d
es
o
r
s
tr
u
ctu
r
es
th
at
ar
e
n
ea
r
b
y
)
,
I
I
I
(
s
p
r
ea
d
with
in
th
e
c
h
est),
to
I
V
(
s
p
r
ea
d
to
d
is
tan
t
o
r
g
a
n
s
)
.
Dete
ctin
g
th
e
d
is
ea
s
e
at
th
e
ea
r
lies
t
s
tag
e
c
an
im
p
r
o
v
e
s
u
r
v
i
v
al
b
u
t
s
in
ce
m
o
s
t
p
atien
ts
ar
e
asy
m
p
to
m
a
tic
b
y
th
e
tim
e
th
e
d
is
ea
s
e
d
ev
elo
p
s
,
th
e
cu
r
r
en
tly
av
ailab
le
tr
ea
tm
en
t a
p
p
r
o
ac
h
e
s
ar
e
in
ef
f
ec
tiv
e.
I
n
id
en
tif
y
in
g
a
n
d
d
ia
g
n
o
s
in
g
NSC
L
C
,
v
ar
io
u
s
im
ag
in
g
tech
n
iq
u
es
ar
e
u
s
u
ally
u
s
e
d
s
u
ch
as,
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
(
CT
)
,
p
o
s
itro
n
em
is
s
io
n
to
m
o
g
r
a
p
h
y
(
PET
)
,
an
d
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
(
MRI)
[
4
]
.
C
T
s
ca
n
s
p
lay
th
e
k
ey
r
o
le
in
th
e
d
iag
n
o
s
is
an
d
s
tag
in
g
o
f
lu
n
g
ca
n
ce
r
,
wh
ich
is
ch
ar
ac
ter
ized
b
y
h
ig
h
-
r
eso
lu
tio
n
im
ag
es,
wh
ich
ca
n
b
e
u
s
ed
to
p
r
o
v
id
e
an
an
ato
m
ical
u
n
d
er
s
tan
d
in
g
o
f
lu
n
g
tu
m
o
r
s
.
T
r
ad
itio
n
al
r
ad
io
l
o
g
ical
ev
alu
atio
n
o
f
C
T
im
ag
es
is
lim
ited
ev
en
th
o
u
g
h
it
is
wid
ely
u
s
ed
.
T
h
e
p
r
o
ce
d
u
r
e
is
v
er
y
m
u
ch
r
elian
t
o
n
t
h
e
ex
p
e
r
ien
ce
o
f
th
e
r
ad
io
l
o
g
is
t
an
d
t
en
d
s
to
c
o
n
ce
n
tr
ate
o
n
th
e
v
is
u
al
f
ea
tu
r
es
o
f
t
h
e
tu
m
o
r
s
s
u
ch
as
s
ize,
s
h
ap
e
,
an
d
lo
ca
tio
n
.
Su
c
h
e
x
am
in
ati
o
n
s
ar
e
s
u
b
jectiv
e
a
n
d
th
u
s
ca
u
s
in
g
d
is
p
ar
ity
in
d
iag
n
o
s
in
g
an
d
s
tag
in
g
.
I
n
a
d
d
itio
n
,
tr
a
d
itio
n
al
im
ag
i
n
g
p
r
o
ce
d
u
r
es
m
ig
h
t
n
o
t
b
e
ab
l
e
to
id
en
tif
y
s
lig
h
t
v
ar
iatio
n
b
etwe
en
ea
r
ly
an
d
la
te
s
tag
es o
f
NSC
L
C
d
ec
r
ea
s
in
g
th
e
p
o
s
s
ib
ilit
y
o
f
ea
r
ly
tr
ea
t
m
en
t.
I
n
o
r
d
e
r
to
ad
d
r
ess
th
ese
co
n
s
tr
ain
ts
,
r
ad
io
m
ics
h
as
b
ec
o
m
e
an
ef
f
ec
tiv
e
m
eth
o
d
in
ca
n
c
er
s
tu
d
ies,
allo
win
g
th
e
id
en
tific
atio
n
o
f
q
u
an
titativ
e
ch
ar
ac
ter
is
tics
o
f
m
ed
ical
im
ag
es,
wh
ich
ca
n
b
e
u
s
ed
to
d
etec
t
tu
m
o
r
p
r
o
p
er
ties
th
at
ar
e
u
n
o
b
s
er
v
ab
le
b
y
th
e
h
u
m
a
n
ey
e
[
5
]
.
T
h
ese
ch
ar
ac
ter
is
tics
,
in
clu
d
in
g
tex
tu
r
e,
s
h
ap
e
,
an
d
in
ten
s
ity
,
r
ep
r
esen
t
a
c
o
m
p
lete
tu
m
o
r
p
h
en
o
t
y
p
e
a
n
d
h
av
e
d
em
o
n
s
tr
ated
p
o
ten
tial
in
a
wid
e
r
a
n
g
e
o
f
o
n
co
lo
g
ical
u
s
es,
s
u
ch
as
tu
m
o
r
ch
ar
ac
te
r
izatio
n
,
p
r
ed
ict
io
n
o
f
tr
ea
tm
e
n
t
r
esp
o
n
s
e,
a
n
d
p
r
o
g
n
o
s
ticatio
n
o
f
o
u
tc
o
m
e
[
6
]
−
[
8
]
.
R
ad
io
m
i
cs
ca
n
b
e
u
s
ed
to
id
e
n
tify
p
atter
n
s
an
d
co
r
r
elatio
n
s
th
at
ca
n
g
u
id
e
clin
ical
d
ec
is
io
n
-
m
ak
in
g
b
y
ap
p
l
y
in
g
m
ac
h
in
e
lear
n
in
g
(
ML
)
m
o
d
el
s
to
s
u
ch
f
ea
tu
r
es
an
d
h
elp
to
p
r
ed
ict
th
e
s
tag
e
o
f
NSC
L
C
ea
r
lier
[
9
]
−
[
1
1
]
.
T
h
is
s
tu
d
y
s
ee
k
s
to
b
u
ild
a
r
a
d
io
m
ics
m
o
d
el
to
d
if
f
e
r
en
tiate
clin
ical
s
tag
e
(
I
to
I
I
I
)
o
f
NS
C
L
C
u
s
in
g
C
T
im
ag
es
b
ased
o
n
th
e
ch
ar
a
cter
is
tics
o
f
tu
m
o
r
s
.
Stag
e
I
V,
wh
ich
is
a
p
r
o
g
r
ess
iv
e
d
is
ea
s
e
wh
er
e
m
etastas
i
s
o
cc
u
r
s
,
is
u
s
u
ally
o
m
itted
s
in
ce
it
h
as
v
er
y
s
p
ec
if
ic
clin
ical
in
d
icato
r
s
th
at
ca
n
b
e
e
asil
y
d
etec
ted
b
y
co
n
v
en
tio
n
al
im
ag
i
n
g
.
B
esid
es,
th
e
tr
ea
tm
e
n
t
o
f
s
tag
e
I
V
i
s
n
o
r
m
ally
p
alliativ
e
u
n
lik
e
s
tag
e
I
to
I
I
I
wh
er
e
tr
ea
tm
en
t
is
m
o
r
e
d
iv
er
s
e.
T
h
er
ef
o
r
e
,
p
r
io
r
itizin
g
s
tag
es
I
to
I
I
I
en
ab
les
th
e
m
o
d
el
to
in
f
o
r
m
d
ec
is
io
n
s
in
ca
s
es wh
er
e
th
e
tr
ea
tm
en
t c
h
o
i
ce
s
ar
e
q
u
ite
d
if
f
e
r
en
t d
e
p
en
d
i
n
g
o
n
th
e
s
tag
e.
A
r
ad
io
m
ics
a
p
p
r
o
ac
h
is
p
r
esen
t
to
class
if
y
t
h
e
NSC
L
C
clin
ical
s
tag
es
u
s
in
g
ML
f
r
o
m
t
wo
p
u
b
licly
av
ailab
le
m
u
lti
-
in
s
titu
tio
n
al
d
atasets
.
T
h
e
r
ad
io
m
ic
f
ea
tu
r
es
wer
e
ex
tr
ac
ted
ac
co
r
d
in
g
to
t
h
e
im
ag
e
b
io
m
a
r
k
er
s
tan
d
ar
d
izatio
n
in
itiativ
e
(
I
B
SI
)
s
tan
d
ar
d
s
[
1
2
]
to
p
r
o
v
id
e
c
o
n
s
is
ten
t
an
d
r
ep
r
o
d
u
cib
le
m
et
h
o
d
o
lo
g
y
ac
r
o
s
s
all
s
tu
d
ies.
T
h
e
h
eter
o
g
en
eity
in
th
e
im
ag
in
g
p
r
o
to
co
ls
is
an
aly
ze
d
an
d
m
itig
ated
u
s
in
g
C
o
m
B
at
h
ar
m
o
n
izatio
n
.
least
ab
s
o
lu
te
s
h
r
in
k
ag
e
a
n
d
s
elec
tio
n
o
p
er
at
o
r
(
L
ASSO
)
an
d
r
ec
u
r
s
iv
e
f
ea
tu
r
e
elim
in
atio
n
(
R
FE
)
wer
e
u
s
ed
to
m
in
im
ize
th
e
d
im
e
n
s
io
n
ality
o
f
th
e
lar
g
e
n
u
m
b
er
o
f
r
a
d
io
m
ic
f
ea
tu
r
es
g
en
er
ate
d
f
r
o
m
e
ac
h
im
ag
e.
Fin
ally
,
s
y
n
th
etic
m
in
o
r
ity
o
v
er
s
am
p
li
n
g
tech
n
i
q
u
e
(
SMOT
E
)
was
u
s
ed
to
ac
co
u
n
t
f
o
r
t
h
e
s
ig
n
if
ic
an
t
class
im
b
alan
ce
in
h
er
en
t
in
th
e
p
atien
t
p
o
p
u
la
tio
n
.
T
h
e
m
u
ltip
le
ML
class
if
ier
s
wer
e
s
y
s
tem
atica
lly
co
m
p
ar
ed
to
d
eter
m
in
e
wh
ich
m
o
d
el
b
est id
en
tifie
d
th
e
co
r
r
ec
t c
lin
ical
s
tag
e
.
T
h
e
k
ey
co
n
t
r
ib
u
tio
n
s
o
f
th
is
wo
r
k
a
r
e
:
First,
d
ev
elo
p
a
r
o
b
u
s
t
r
ad
io
m
ics
p
ip
elin
e
th
at
in
teg
r
ates
s
tan
d
ar
d
ized
f
ea
t
u
r
e
e
x
tr
ac
tio
n
,
b
atch
h
ar
m
o
n
izatio
n
,
an
d
m
o
d
el
v
alid
atio
n
.
Seco
n
d
,
i
d
e
n
tify
a
s
u
b
-
g
r
o
u
p
o
f
d
is
cr
im
in
ativ
e
r
ad
io
m
ic
f
ea
tu
r
es
th
at
ca
n
d
er
iv
e
clin
ically
r
elev
an
t
tu
m
o
r
c
h
ar
ac
ter
is
tics
o
f
NSC
L
C
s
tag
e
.
T
h
ir
d
,
g
i
v
e
a
co
m
p
ar
ativ
e
an
a
ly
s
is
o
f
ML
alg
o
r
ith
m
s
,
wh
ich
p
r
o
v
es
th
at
e
n
s
em
b
le
-
b
ased
m
o
d
els
ar
e
ef
f
ec
tiv
e
in
th
e
co
n
tex
t
o
f
r
a
d
io
m
ics
-
b
a
s
ed
s
tag
e
class
if
icatio
n
.
T
h
e
p
ap
er
is
o
r
g
an
ize
d
as
f
o
llo
ws:
s
ec
tio
n
2
d
escr
ib
es
th
e
d
atasets
,
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
s
elec
tio
n
,
m
o
d
el
tr
ain
i
n
g
an
d
v
alid
atio
n
.
Sectio
n
s
3
d
is
cu
s
s
th
e
r
esu
lts
an
d
co
m
p
ar
e
th
e
m
with
f
in
d
in
g
s
f
r
o
m
p
r
e
v
io
u
s
s
tu
d
ies.
L
astl
y
,
s
ec
tio
n
4
co
n
clu
d
es th
is
s
tu
d
y
.
2.
M
E
T
H
O
D
2
.
1
.
Da
t
a
s
et
I
n
th
is
s
tu
d
y
,
th
e
C
T
im
ag
es
wer
e
o
b
tain
ed
f
r
o
m
two
p
u
b
li
cly
av
ailab
le
m
u
lti
-
in
s
titu
tio
n
al
d
atasets
:
d
ataset
1
–
th
e
NSC
L
C
-
R
ad
io
m
ics
d
ataset
[
1
3
]
,
an
d
d
ata
s
et
2
–
th
e
NSC
L
C
-
r
ad
io
g
en
o
m
ics
d
ataset
[
1
4
]
.
Data
s
et
1
co
n
tain
s
C
T
im
ag
e
s
f
r
o
m
4
2
2
p
atien
ts
,
in
clu
d
in
g
5
1
with
ADC,
1
5
2
with
S
C
C
,
1
1
4
with
L
C
C
,
6
3
as
n
o
t
o
t
h
er
wis
e
s
p
ec
if
ied
(
NOS)
,
an
d
4
2
as
n
o
t
av
ailab
l
e
(
NA)
d
u
e
to
i
n
co
m
p
lete
d
iag
n
o
s
tic
in
f
o
r
m
atio
n
.
E
ac
h
ca
s
e
in
clu
d
es
3
D
g
r
o
s
s
tu
m
o
r
v
o
lu
m
es
(
GT
V)
d
elin
ea
ted
b
y
o
n
co
lo
g
is
ts
alo
n
g
with
ass
o
ciate
d
clin
ical
o
u
tco
m
es.
Fo
llo
win
g
v
is
u
al
i
n
s
p
ec
tio
n
u
s
in
g
th
e
3
D
Sli
ce
r
to
o
l
[
1
5
]
,
a
s
elec
tio
n
o
f
2
5
0
p
atien
t
r
ec
o
r
d
s
with
co
m
p
r
eh
e
n
s
iv
e
clin
ical
d
ata
was
m
ad
e
f
o
r
f
u
r
th
er
an
al
y
s
is
.
T
h
e
d
ataset
2
co
m
p
r
is
es
C
T
im
ag
es
f
r
o
m
1
4
4
NSC
L
C
p
atien
ts
,
in
clu
d
in
g
1
1
2
with
ADC,
2
9
with
SC
C
,
an
d
3
with
NOS.
T
h
e
s
e
im
ag
es
in
clu
d
e
s
eg
m
en
tatio
n
s
in
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
)
f
o
r
m
at
an
d
c
lin
ical
o
u
tco
m
es
in
C
SV
f
iles
.
I
n
itial
s
eg
m
en
tatio
n
s
wer
e
d
o
n
e
with
a
s
em
i
-
au
t
o
m
atic
alg
o
r
ith
m
an
d
later
r
ef
in
ed
b
y
an
ex
p
er
t.
Af
ter
v
er
if
icatio
n
u
s
in
g
th
e
3
D
Sli
ce
r
to
o
l,
1
1
9
p
atien
t r
ec
o
r
d
s
wer
e
ch
o
s
en
f
o
r
a
n
aly
s
is
.
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
C
o
mp
u
ted
to
mo
g
r
a
p
h
y
ima
g
in
g
r
a
d
io
mics:
a
n
o
ve
l a
p
p
r
o
a
c
h
to
ea
r
ly
-
s
ta
g
e
n
o
n
-
s
ma
ll …
(
R
a
viteja
B
a
leka
i
)
2473
T
h
e
s
tu
d
y
ex
clu
d
e
d
p
atien
ts
wh
o
s
e
clin
ical
s
tag
e
lab
els
we
r
e
lack
in
g
o
r
am
b
ig
u
o
u
s
,
h
ad
in
co
m
p
lete
o
r
co
r
r
u
p
te
d
C
T
im
ag
es,
a
n
d
t
u
m
o
r
s
eg
m
en
tatio
n
m
ask
s
wer
e
in
co
n
s
is
ten
t
o
r
ab
s
en
t
to
g
u
ar
an
tee
q
u
ality
an
d
r
ep
r
o
d
u
cib
ilit
y
o
f
d
ata.
Af
ter
th
is
p
r
o
ce
s
s
o
f
in
clu
s
io
n
a
n
d
ex
clu
s
io
n
,
a
t
o
tal
o
f
3
6
9
p
atien
t
r
ec
o
r
d
s
wer
e
r
etain
ed
f
r
o
m
two
d
atasets
to
co
n
tin
u
e
with
th
e
s
u
b
s
eq
u
en
t
an
aly
s
es.
I
n
t
h
is
r
esear
ch
,
C
T
s
ca
n
s
wer
e
ac
q
u
ir
e
d
f
r
o
m
d
if
f
er
e
n
t
in
s
titu
tio
n
s
,
th
er
ef
o
r
e
,
th
er
e
ar
e
v
a
r
iatio
n
s
in
th
e
im
a
g
e
ac
q
u
is
itio
n
p
r
o
to
co
ls
d
u
e
t
o
th
e
d
if
f
er
en
ce
s
o
f
th
e
s
ca
n
n
er
m
an
u
f
ac
tu
r
e
r
s
an
d
th
e
im
a
g
e
a
cq
u
is
itio
n
p
ar
am
eter
s
.
I
n
o
r
d
er
to
ev
alu
ate
th
is
h
eter
o
g
en
eity
s
cien
tific
ally
,
DI
C
OM
m
etad
ata
wer
e
an
al
y
ze
d
with
th
e
h
elp
o
f
3
D
Sli
ce
r
,
th
e
d
etails
ar
e
p
r
o
v
id
e
d
in
T
a
b
le
1
.
T
ab
le
1
.
I
m
ag
in
g
p
ar
am
eter
s
o
f
ea
ch
C
T
s
ca
n
D
a
t
a
s
e
t
P
i
x
e
l
s
p
a
c
i
n
g
[
x
,
y
]
i
n
m
m
S
l
i
c
e
t
h
i
c
k
n
e
ss
[
z
]
i
n
mm
M
a
t
r
i
x
si
z
e
[
x
,
y
]
D
a
t
a
s
e
t
1
[
0
.
9
7
6
5
6
2
5
0
,
0
.
9
7
6
5
6
2
5
]
2
.
9
9
[
5
1
2
×
5
1
2
]
D
a
t
a
s
e
t
2
[
0
.
5
9
7
6
5
6
,
0
.
9
7
6
5
6
]
[
0
.
6
2
5
−
2
.
5
]
[
5
1
2
×
5
1
2
]
2
.
2
.
F
e
a
t
ure
ex
t
r
a
ct
io
n a
nd
ba
t
ch
ha
rm
o
niza
t
io
n
T
h
e
Py
R
ad
io
m
ics
lib
r
ar
y
[
1
6
]
,
in
teg
r
ated
in
to
th
e
3
D
Sli
ce
r
,
an
o
p
e
n
-
s
o
u
r
ce
s
o
f
twar
e
f
o
r
v
is
u
aliza
tio
n
an
d
im
a
g
e
an
aly
s
is
,
was
u
s
ed
to
ex
tr
ac
t
r
ad
io
m
ic
f
ea
tu
r
es
f
r
o
m
th
e
GT
V,
wh
ich
was
r
esam
p
led
to
a
r
eso
lu
tio
n
o
f
1
×1
×1
m
m
3
.
T
h
is
p
r
ep
r
o
ce
s
s
in
g
s
tep
was
ca
r
r
ied
o
u
t
to
p
r
ep
ar
e
t
h
e
s
am
p
le
co
n
s
is
ten
t,
an
d
th
e
f
o
llo
win
g
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
m
o
d
elin
g
s
tep
s
m
o
r
e
r
e
liab
le.
T
h
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
was
d
o
n
e
b
ased
o
n
th
e
I
B
SI
f
ea
tu
r
e
d
ef
in
itio
n
s
.
1
0
7
r
ad
i
o
m
ic
f
ea
tu
r
es
is
ex
tr
ac
ted
f
o
r
ea
ch
p
atien
t,
wh
ich
wer
e
g
r
o
u
p
e
d
in
t
o
th
r
ee
ca
teg
o
r
ies:
i)
m
o
r
p
h
o
lo
g
ical
(
1
4
f
ea
tu
r
es),
ii)
f
ir
s
t
-
o
r
d
er
s
tatis
tical
(
1
8
f
ea
tu
r
e
s
)
,
an
d
iii)
tex
tu
r
e
(
7
5
f
ea
t
u
r
es).
All th
e
r
a
d
io
m
ic
f
ea
tu
r
es we
r
e
z
-
s
co
r
e
n
o
r
m
alize
d
to
g
et
a
n
o
r
m
al
d
is
tr
ib
u
tio
n
.
R
ad
io
m
ic
f
ea
tu
r
es
ca
n
b
e
a
f
f
ec
ted
b
y
th
e
s
ca
n
n
er
ty
p
e
,
ac
q
u
is
itio
n
s
eq
u
en
ce
s
an
d
r
ec
o
n
s
tr
u
ctio
n
p
ar
am
eter
s
,
p
a
r
ticu
lar
ly
in
m
u
lti
-
s
ite
d
ata.
I
n
th
is
s
tu
d
y
,
C
T
im
ag
es
wer
e
ac
q
u
ir
e
d
f
r
o
m
d
i
f
f
er
en
t
in
s
titu
tio
n
s
an
d
d
if
f
e
r
en
t
s
ca
n
n
er
m
o
d
els,
an
d
s
o
,
b
atch
ef
f
ec
t
was
tak
en
in
to
ac
co
u
n
t
p
r
io
r
t
o
b
u
il
d
in
g
a
m
o
d
el
[
1
7
]
.
Prin
cip
al
co
m
p
o
n
en
t
an
aly
s
i
s
(
PC
A
)
an
d
t
-
d
is
tr
ib
u
ted
s
to
ch
asti
c
n
eig
h
b
o
r
em
b
e
d
d
i
n
g
(
t
-
SNE
)
wer
e
p
er
f
o
r
m
ed
o
n
s
tan
d
a
r
d
ized
f
ea
tu
r
es
an
d
s
h
o
wed
clu
s
ter
in
g
b
y
d
ataset,
s
u
g
g
esti
n
g
th
at
th
er
e
was
a
p
r
esen
ce
o
f
s
ca
n
n
er
-
a
n
d
s
ite
-
s
p
ec
if
ic
v
a
r
iatio
n
s
.
T
o
elim
in
ate
th
e
ef
f
e
cts,
C
o
m
B
at
h
ar
m
o
n
izatio
n
was
p
er
f
o
r
m
ed
,
with
d
ataset
o
r
ig
in
as a
b
atch
c
o
v
ar
iate.
Af
ter
h
ar
m
o
n
izatio
n
,
th
e
PC
A
an
d
t
-
SNE
clu
s
ter
in
g
v
alu
es we
r
e
lo
wer
an
d
f
ea
tu
r
es
wer
e
b
etter
d
is
tr
ib
u
t
ed
,
co
n
f
ir
m
in
g
t
h
e
s
u
cc
ess
f
u
l
r
em
o
v
al
o
f
n
o
n
-
b
io
lo
g
ical
v
ar
iab
ilit
y
,
wh
ile
p
r
eser
v
in
g
t
u
m
o
r
-
s
p
ec
if
ic
in
f
o
r
m
atio
n
.
T
h
is
r
esu
lted
in
b
ette
r
an
d
s
tr
o
n
g
er
r
ad
io
m
ic
f
ea
tu
r
es f
o
r
m
o
d
elin
g
.
2
.
3
.
F
e
a
t
ure
s
elec
t
io
n
T
h
is
s
tep
is
a
cr
u
cial
in
s
elec
tin
g
r
ad
io
m
ic
f
ea
t
u
r
es
to
p
r
e
d
ict
th
e
s
tag
es
o
f
NSC
L
C
in
C
T
s
ca
n
s
.
I
n
th
is
r
esear
ch
,
th
e
R
FE
[
1
8
]
an
d
L
ASSO
[
1
9
]
,
[
2
0
]
m
eth
o
d
s
wer
e
em
p
lo
y
ed
f
o
r
f
ea
tu
r
e
s
elec
tio
n
.
R
FE
was
u
s
ed
as
a
wr
ap
p
er
-
b
ased
f
ea
t
u
r
e
s
elec
tio
n
tech
n
iq
u
e.
R
FE
r
em
o
v
es
f
ea
tu
r
es
o
n
e
at
a
tim
e
an
d
f
its
th
e
m
o
d
el
with
th
e
r
em
ai
n
in
g
f
ea
tu
r
es,
an
d
th
e
im
p
o
r
tan
ce
o
f
th
e
f
e
atu
r
es
is
m
ea
s
u
r
ed
ac
co
r
d
in
g
to
a
s
p
ec
if
ic
ML
m
o
d
el.
T
h
is
p
r
o
ce
s
s
is
r
ep
ea
ted
u
n
til
th
e
n
u
m
b
er
o
f
f
ea
tu
r
e
s
is
eq
u
al
to
th
e
d
esire
d
n
u
m
b
er
o
f
f
ea
tu
r
es
to
b
e
r
etain
ed
,
an
d
th
e
least
im
p
o
r
tan
t
f
ea
tu
r
es
ar
e
d
r
o
p
p
ed
.
M
o
r
eo
v
e
r
,
L
ASSO
u
s
es
L
1
r
e
g
u
lar
izatio
n
,
w
h
ich
s
h
r
in
k
s
th
e
ab
s
o
lu
te
v
alu
es
o
f
f
ea
tu
r
e
co
ef
f
icien
ts
,
th
u
s
s
ettin
g
less
im
p
o
r
tan
t
f
ea
t
u
r
es'
co
ef
f
icien
ts
to
ze
r
o
.
T
h
is
elim
in
ates
n
o
n
-
im
p
o
r
tan
t
o
r
r
ed
u
n
d
a
n
t
f
ea
t
u
r
es,
a
n
d
k
e
ep
s
o
n
ly
th
o
s
e
th
at
p
lay
a
r
o
le
in
t
h
e
p
r
ed
ictiv
e
m
o
d
el.
T
h
r
o
u
g
h
f
ea
tu
r
e
r
e
d
u
c
tio
n
,
L
ASSO
r
ed
u
ce
s
m
o
d
el
c
o
m
p
lex
ity
,
en
h
a
n
ce
s
in
ter
p
r
etab
ilit
y
an
d
r
ed
u
ce
s
o
v
er
f
itti
n
g
.
T
h
e
ex
p
r
ess
io
n
o
f
L
ass
o
is
s
h
o
wn
in
(
1
)
.
(
)
=
∑
(
(
)
−
∑
=
0
(
)
)
2
=
1
+
∑
|
|
=
0
(
1
)
W
h
er
e,
is
th
e
n
u
m
b
er
o
f
o
b
s
er
v
atio
n
s
,
is
th
e
n
u
m
b
er
o
f
f
ea
tu
r
es,
(
)
is
i
th
s
am
p
le,
(
)
is
o
b
s
er
v
ed
o
u
tp
u
t
o
f
th
e
i
th
s
am
p
le,
θ
is
th
e
r
eg
r
ess
io
n
co
e
f
f
icien
t
,
an
d
∑
|
|
=
0
is
r
eg
u
la
r
izatio
n
ter
m
with
th
e
r
eg
u
lar
izatio
n
p
ar
am
eter
λ
.
B
y
in
co
r
p
o
r
atin
g
L
ASSO
an
d
R
FE
in
to
t
h
e
f
ea
t
u
r
e
s
elec
tio
n
p
r
o
ce
s
s
,
th
is
s
tu
d
y
aim
s
to
e
n
h
an
ce
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
p
r
ed
ictiv
e
m
o
d
el.
I
t
also
s
ee
k
s
to
im
p
r
o
v
e
in
ter
p
r
etab
ilit
y
b
y
f
o
c
u
s
in
g
o
n
th
e
m
o
s
t
d
is
cr
im
in
ativ
e
r
ad
i
o
m
ics
f
ea
t
u
r
es.
I
n
ad
d
itio
n
,
t
h
is
ap
p
r
o
ac
h
h
elp
s
m
itig
ate
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
an
d
co
m
p
u
tatio
n
al
c
o
m
p
lex
ity
.
2
.
4
.
M
o
del
t
ra
ini
ng
a
nd
v
a
li
da
t
io
n
T
h
is
s
tu
d
y
aim
ed
to
cr
ea
te
a
f
r
am
ewo
r
k
o
f
class
if
icatio
n
in
p
r
ed
ictin
g
t
h
e
clin
ical
s
tag
es
o
f
NSC
L
C
.
T
h
er
e
was
a
s
ig
n
if
ican
t
im
b
a
lan
ce
in
class
es
in
th
e
d
ataset
with
s
tag
e
I
(
n
=
1
1
8
)
,
s
tag
e
I
I
(
n
=4
9
)
,
a
n
d
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.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
4
7
1
-
2
4
8
3
2474
s
tag
e
I
I
I
(
n
=2
0
2
)
ca
s
es.
T
o
r
e
d
u
ce
th
e
p
o
s
s
ib
ilit
y
o
f
in
f
o
r
m
atio
n
leak
ag
e
,
f
ea
tu
r
e
s
elec
tio
n
b
ased
o
n
R
FE
an
d
L
ASSO
an
d
o
v
er
s
am
p
lin
g
wa
s
o
n
ly
u
s
ed
o
n
th
e
tr
ai
n
in
g
d
a
ta
in
ea
ch
f
o
ld
o
f
th
e
cr
o
s
s
-
v
a
lid
atio
n
.
T
h
e
class
im
b
alan
ce
was
a
d
d
r
ess
ed
b
y
u
s
in
g
th
e
SMOT
E
[
2
1
]
to
th
e
tr
ain
in
g
d
ataset,
i
n
wh
ic
h
th
e
s
y
n
th
etic
m
in
o
r
ity
class
ex
am
p
les
wer
e
g
en
er
ated
u
s
in
g
th
e
k
-
n
ea
r
est
n
eig
h
b
o
r
alg
o
r
ith
m
a
p
p
r
o
ac
h
,
r
esu
ltin
g
in
a
b
etter
class
b
alan
ce
an
d
b
etter
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
m
o
d
els.
R
an
d
o
m
s
elec
tio
n
o
f
th
e
s
u
b
s
et
o
f
tr
ai
n
in
g
an
d
test
d
ata
was
d
o
n
e
with
a
n
8
0
-
2
0
s
p
lit
an
d
th
e
wh
o
le
s
er
ies
o
f
C
T
im
a
g
es
o
f
ea
ch
p
atien
t
was
u
s
ed
.
T
h
e
m
o
d
els
wer
e
tr
ain
ed
with
f
iv
e
d
if
f
er
e
n
t
ML
class
if
ier
s
;
lo
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
,
d
ec
is
io
n
tr
ee
(
DT
)
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM)
,
r
an
d
o
m
f
o
r
est
(
R
F)
,
an
d
e
x
tr
em
e
g
r
ad
ien
t
b
o
o
s
tin
g
(
XGBo
o
s
t)
.
T
h
e
ev
alu
atio
n
with
m
u
ltip
le
class
if
ier
s
is
ess
en
tia
l
in
r
a
d
io
m
ics
s
tu
d
ies,
s
in
ce
class
if
ier
s
ca
n
ac
h
iev
e
d
if
f
er
e
n
t
p
er
f
o
r
m
a
n
ce
s
in
ter
m
s
o
f
f
ea
tu
r
es,
class
im
b
alan
ce
,
an
d
co
m
p
le
x
ity
o
f
th
e
d
ata
ass
o
ciatio
n
.
T
o
ev
alu
ate
th
e
r
eliab
ilit
y
an
d
r
o
b
u
s
tn
ess
o
f
th
e
m
o
d
els,
th
e
av
er
ag
e
p
er
f
o
r
m
an
ce
o
f
ac
c
u
r
ac
y
,
s
en
s
itiv
ity
,
s
p
ec
if
icity
,
F1
-
s
co
r
e
,
an
d
th
e
ar
ea
u
n
d
er
t
h
e
r
ec
eiv
er
o
p
e
r
atin
g
c
h
ar
ac
ter
is
tic
(
AUROC
)
ac
r
o
s
s
th
e
f
o
ld
s
was c
alcu
lated
.
3.
RE
SU
L
T
S
AND
D
I
SCU
SS
I
O
N
3
.
1
.
Resul
t
s
A
to
tal
o
f
3
6
9
p
atien
ts
with
NSC
L
C
wer
e
s
tu
d
ied
.
T
h
e
av
er
ag
e
ag
e
o
f
th
e
p
atien
ts
was
6
8
.
8
0
y
ea
r
s
,
an
d
2
5
6
(
6
9
.
3
7
%)
o
f
th
e
p
a
tien
ts
wer
e
m
ale.
His
to
lo
g
ic
al
s
u
b
ty
p
es
o
f
th
e
p
atien
ts
wer
e:
1
3
5
(
ADC),
1
4
5
(
SC
C
)
,
an
d
8
9
(
L
C
C
)
.
R
e
g
ar
d
in
g
clin
ical
s
tag
es,
th
er
e
wer
e
1
1
8
p
atien
ts
(
3
1
.
9
7
%)
in
s
tag
e
I
,
4
9
p
atien
ts
(
1
3
.
2
7
%)
in
s
tag
e
I
I
,
a
n
d
2
0
2
p
atien
ts
(
5
4
.
7
4
%)
in
s
tag
e
I
I
I
.
T
a
b
le
s
2
an
d
3
p
r
o
v
id
es
s
u
m
m
ar
y
o
u
tlin
in
g
d
em
o
g
r
a
p
h
ic
an
d
clin
ical
in
f
o
r
m
atio
n
ac
r
o
s
s
two
d
is
tin
ct
d
atasets
.
T
ab
le
2
.
Dem
o
g
r
ap
h
ic
an
d
clin
ical
ch
ar
ac
ter
is
tics
o
f
th
e
NSC
L
C
-
R
ad
io
m
ics d
atase
t
S
u
b
j
e
c
t
c
h
a
r
a
c
t
e
r
i
st
i
c
s
ADC
S
C
C
LC
C
N
u
mb
e
r
o
f
su
b
j
e
c
t
s
42
1
1
9
89
G
e
n
d
e
r
m
a
l
e
26
84
55
G
e
n
d
e
r
f
e
m
a
l
e
16
35
34
A
g
e
m
e
a
n
a
g
e
(
y
e
a
r
s)
6
7
.
3
1
6
9
.
7
4
6
7
.
3
2
A
g
e
r
a
n
g
e
4
5
.
7
2
-
8
5
.
6
0
3
3
.
6
8
-
8
8
.
3
8
4
6
.
3
4
-
9
1
.
7
0
S
t
a
g
e
I
9
18
13
S
t
a
g
e
II
3
20
4
S
t
a
g
e
I
I
I
30
81
72
T
ab
le
3
.
Dem
o
g
r
ap
h
ic
an
d
clin
ical
ch
ar
ac
ter
is
tics
o
f
th
e
NSC
L
C
-
R
ad
io
g
en
o
m
ics d
ataset
S
u
b
j
e
c
t
c
h
a
r
a
c
t
e
r
i
st
i
c
s
ADC
S
C
C
N
u
mb
e
r
o
f
su
b
j
e
c
t
s
93
26
G
e
n
d
e
r
m
a
l
e
69
22
G
e
n
d
e
r
f
e
m
a
l
e
24
4
A
g
e
m
e
a
n
a
g
e
(
y
e
a
r
s)
6
8
.
9
6
7
0
.
6
9
A
g
e
r
a
n
g
e
43
-
87
57
-
83
S
t
a
g
e
I
62
16
S
t
a
g
e
II
16
6
S
t
a
g
e
I
I
I
15
4
E
ac
h
tu
m
o
r
v
o
lu
m
e
u
n
d
er
we
n
t
ex
ten
s
iv
e
ch
ar
ac
ter
izatio
n
b
y
ex
tr
ac
tin
g
a
co
m
p
r
eh
e
n
s
iv
e
s
et
o
f
1
0
7
f
ea
tu
r
es.
T
h
ese
q
u
an
titati
v
e
m
ea
s
u
r
es
ca
n
b
e
ca
te
g
o
r
iz
ed
in
to
t
h
r
ee
p
r
im
ar
y
g
r
o
u
p
s
:
i)
m
o
r
p
h
o
lo
g
ical
(
s
h
ap
e)
attr
ib
u
tes,
ii)
s
tatis
tic
al
attr
ib
u
tes
o
f
th
e
f
ir
s
t
o
r
d
er
,
an
d
iii)
tex
tu
r
e
attr
ib
u
tes.
T
h
e
tex
tu
r
e
f
ea
tu
r
es
co
n
s
is
t
o
f
2
4
t
h
e
g
r
a
y
lev
el
c
o
-
o
cc
u
r
r
e
n
ce
m
atr
ix
(
GL
C
M)
,
1
4
g
r
a
y
lev
el
d
ep
en
d
en
ce
m
at
r
ix
(
GL
DM
)
,
1
6
th
e
g
r
ay
lev
el
r
u
n
le
n
g
th
m
at
r
ix
(
GL
R
L
M)
,
1
6
th
e
g
r
ay
lev
el
s
i
ze
zo
n
e
m
atr
i
x
(
GL
SZM
)
,
an
d
5
th
e
n
eig
h
b
o
r
in
g
g
r
ay
to
n
e
d
if
f
e
r
en
ce
m
atr
i
x
(
N
GT
DM
)
f
ea
tu
r
es.
T
h
e
an
aly
s
is
o
f
s
tan
d
ar
d
ized
r
ad
io
m
ic
f
ea
tu
r
es
with
PC
A
an
d
t
-
SNE
was
d
o
n
e
to
d
et
er
m
in
e
th
e
p
r
esen
ce
o
f
b
atch
-
r
elate
d
v
a
r
iab
ilit
y
.
T
h
e
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
m
eth
o
d
s
wer
e
u
s
ed
to
v
is
u
alize
th
e
d
is
tr
ib
u
tio
n
s
o
f
f
ea
tu
r
es
an
d
d
eter
m
in
e
p
o
s
s
ib
le
clu
s
ter
tr
en
d
s
r
elate
d
to
th
e
s
o
u
r
ce
o
f
d
at
a.
Fig
u
r
e
1
clea
r
ly
s
h
o
ws
d
ataset
-
s
p
ec
if
ic
clu
s
ter
in
g
,
wh
ich
is
s
tr
o
n
g
e
v
id
en
c
e
o
f
lar
g
e
b
atch
ef
f
ec
ts
o
f
s
ca
n
n
er
s
,
ac
q
u
is
itio
n
p
r
o
to
co
ls
,
an
d
in
s
titu
tio
n
al
p
r
a
ctice
.
B
esid
es
v
is
u
al
ex
am
in
atio
n
,
th
e
s
tatis
tical
s
u
m
m
ar
ies
(
in
b
atch
)
o
f
m
ea
n
,
m
ed
ian
,
an
d
s
tan
d
ar
d
d
ev
iatio
n
o
f
f
e
atu
r
es we
r
e
ca
lcu
lated
t
o
m
ea
s
u
r
e
th
e
i
n
ter
-
d
ataset
v
ar
iatio
n
.
T
o
r
ed
u
ce
th
ese
n
o
n
-
b
i
o
lo
g
ic
al
v
ar
iatio
n
s
th
e
C
o
m
B
at
h
ar
m
o
n
izatio
n
alg
o
r
ith
m
was
u
s
ed
o
n
th
e
r
ad
io
m
ic
f
ea
tu
r
e
m
atr
i
x
.
C
o
m
B
at
is
an
em
p
ir
ical
B
ay
es
ap
p
r
o
ac
h
th
at
is
co
m
m
o
n
ly
u
s
ed
in
r
ad
io
m
ics
liter
atu
r
e
to
co
r
r
ec
t
b
atch
e
f
f
ec
t
wh
ils
t
m
ain
tain
in
g
b
io
lo
g
ically
m
ea
n
in
g
f
u
l
s
ig
n
al.
I
n
t
h
is
wo
r
k
,
th
e
b
atch
co
v
ar
iate
was
d
ataset
o
r
ig
in
,
an
d
th
e
h
a
r
m
o
n
izatio
n
was
ac
h
iev
ed
with
th
e
u
s
e
o
f
th
e
p
y
co
m
b
at
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
C
o
mp
u
ted
to
mo
g
r
a
p
h
y
ima
g
in
g
r
a
d
io
mics:
a
n
o
ve
l a
p
p
r
o
a
c
h
to
ea
r
ly
-
s
ta
g
e
n
o
n
-
s
ma
ll …
(
R
a
viteja
B
a
leka
i
)
2475
im
p
lem
en
tatio
n
at
h
ttp
s
://g
ith
u
b
.
co
m
/J
f
o
r
tin
1
/C
o
m
B
atHa
r
m
o
n
izatio
n
.
T
h
is
alg
o
r
ith
m
was
u
s
ed
o
n
th
e
tr
an
s
p
o
s
ed
f
ea
tu
r
e
m
atr
ix
to
m
ak
e
s
u
r
e
th
at
th
er
e
was p
r
o
p
e
r
ad
ju
s
tm
en
t a
m
o
n
g
s
t f
ea
tu
r
es.
Af
ter
h
ar
m
o
n
izatio
n
,
th
e
b
atc
h
co
r
r
ec
tio
n
was
ag
ain
test
ed
b
y
PC
A
an
d
t
-
SNE
an
aly
s
i
s
.
Fig
u
r
e
2
s
h
o
ws th
at
d
ataset
s
p
ec
if
ic
clu
s
ter
in
g
was sig
n
if
ican
tly
less
,
an
d
th
er
e
was le
s
s
o
v
er
lap
in
f
ea
tu
r
e
d
is
tr
ib
u
tio
n
s
b
etwe
en
b
atch
es.
T
h
ese
f
in
d
in
g
s
s
u
p
p
o
r
t
th
at
C
o
m
B
at
w
as
an
ef
f
ec
tiv
e
way
o
f
r
ed
u
cin
g
s
ca
n
n
er
-
a
n
d
s
ite
-
ef
f
ec
ts
with
o
u
t
lo
s
s
o
f
d
i
s
cr
im
in
ativ
e
in
f
o
r
m
atio
n
ab
o
u
t
tu
m
o
r
s
.
Su
ch
h
a
r
m
o
n
izatio
n
co
n
tr
ib
u
ted
to
th
e
s
tr
en
g
th
an
d
r
ep
ea
tab
ilit
y
o
f
th
e
r
ad
io
m
ic
f
ea
t
u
r
es th
at
wer
e
t
o
b
e
f
u
r
th
er
m
o
d
eled
.
Fig
u
r
e
1
.
T
h
e
PC
A
an
d
t
-
SNE
s
ca
tter
p
lo
t illu
s
tr
ate
th
e
d
is
tr
ib
u
tio
n
o
f
th
e
th
r
ee
s
tag
es c
lass
d
ataset
Fig
u
r
e
2
.
T
h
e
PC
A
an
d
t
-
SNE
s
ca
tter
p
lo
t a
f
ter
h
ar
m
o
n
izatio
n
T
o
id
en
tify
th
e
m
o
s
t
d
is
tin
ctiv
e
s
u
b
s
et
o
f
f
ea
tu
r
es
an
d
e
n
h
an
ce
m
o
d
el
in
ter
p
r
etab
ilit
y
,
\
a
f
ea
tu
r
e
s
elec
tio
n
s
tr
ateg
y
u
s
in
g
R
FE
an
d
L
ASSO
was
im
p
lem
en
te
d
.
Fo
r
R
FE,
th
e
f
ea
tu
r
e
co
u
n
t
was
s
et
to
1
5
wh
ile
L
ASSO
u
s
ed
a
r
eg
u
lar
izatio
n
p
ar
am
eter
λ
o
f
0
.
1
.
T
h
e
n
u
m
b
er
o
f
f
ea
t
u
r
es
to
r
etain
b
y
R
FE
(
n
=1
5
)
[
2
2
]
,
was
s
et
b
ased
o
n
p
r
ev
io
u
s
r
ad
io
m
ics
s
tu
d
ies
r
ep
o
r
tin
g
th
at
s
m
all
f
ea
tu
r
e
s
ets
im
p
r
o
v
e
g
en
e
r
aliza
tio
n
wh
ile
n
o
t
co
m
p
r
o
m
is
in
g
d
is
cr
im
in
atio
n
in
m
ed
ical
class
if
icatio
n
task
s
with
m
u
ltip
le
cl
ass
es.
T
h
e
in
itia
l
ex
p
er
im
en
tatio
n
s
h
o
wed
th
at
s
elec
tin
g
f
ewe
r
th
an
1
0
f
e
atu
r
es
led
to
u
n
s
tab
le
class
i
f
icatio
n
wh
ile
o
v
er
2
0
f
ea
tu
r
es
led
to
in
cr
ea
s
ed
o
v
er
f
itti
n
g
with
o
u
t
p
er
f
o
r
m
a
n
c
e
g
ain
.
Similar
ly
,
th
e
L
ASSO
p
en
alty
p
ar
am
eter
(λ
=0
.
1
)
[
2
3
]
was
d
eter
m
in
ed
em
p
ir
ically
o
n
th
e
s
tab
ilit
y
o
f
th
e
f
ea
tu
r
es
s
elec
ted
ac
r
o
s
s
f
o
ld
s
an
d
co
n
g
r
u
en
c
e
with
r
ad
io
m
ics
-
b
ased
NSC
L
C
ex
p
er
im
en
ts
p
u
b
lis
h
ed
in
th
e
liter
atu
r
e.
T
h
ese
p
ar
am
eter
s
o
f
f
er
a
g
o
o
d
co
m
p
r
o
m
is
e
b
etwe
en
f
ea
tu
r
e
s
elec
tio
n
an
d
m
o
d
el
co
m
p
l
ex
ity
,
av
o
id
in
g
o
v
er
f
itti
n
g
a
n
d
e
n
s
u
r
in
g
b
o
th
in
ter
p
r
etab
ilit
y
an
d
h
ig
h
p
er
f
o
r
m
an
ce
.
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.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
4
7
1
-
2
4
8
3
2476
So
m
e
o
f
th
e
r
ad
i
o
m
ic
f
ea
tu
r
es
th
at
ar
e
m
o
s
t
co
m
m
o
n
ly
ch
o
s
en
ar
e
clin
ically
r
elev
a
n
t.
Sh
ap
e
f
ea
tu
r
es,
in
clu
d
in
g
Ma
jo
r
Ax
is
L
en
g
th
,
Flatn
ess
,
an
d
Sp
h
er
ic
ity
,
ar
e
in
d
icato
r
s
o
f
tu
m
o
r
s
ize,
elo
n
g
atio
n
,
an
d
g
eo
m
etr
ical
ir
r
eg
u
lar
ity
,
wh
ic
h
h
av
e
b
ee
n
f
o
u
n
d
to
r
is
e
as
d
is
ea
s
e
s
tag
e
p
r
o
g
r
ess
es.
I
m
c1
,
zo
n
e
en
tr
o
p
y
,
an
d
Gr
ay
L
ev
elNo
n
Un
if
o
r
m
ity
ar
e
tex
tu
r
e
m
ea
s
u
r
es
u
s
ed
to
m
e
asu
r
e
in
tr
atu
m
o
r
al
h
eter
o
g
en
e
ity
,
wh
ich
h
as
b
ee
n
id
en
tifie
d
to
b
e
ass
o
ciate
d
w
ith
ag
g
r
ess
iv
e
tu
m
o
r
b
eh
av
i
o
r
an
d
n
o
d
al
in
v
o
lv
em
e
n
t.
T
h
e
s
im
ilar
ity
in
th
e
ch
ar
ac
ter
is
tics
o
f
th
ese
f
ea
tu
r
es
in
b
o
th
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
in
d
icate
s
th
at
th
e
m
o
d
el
is
ca
p
tu
r
i
n
g
b
io
lo
g
ically
s
ig
n
if
ica
n
t
f
ea
tu
r
es,
as
o
p
p
o
s
ed
to
th
e
n
o
is
e
f
o
u
n
d
in
th
e
d
ataset,
an
d
en
h
an
ce
s
clin
ical
in
ter
p
r
etab
ilit
y
.
T
h
is
s
elec
tio
n
p
r
o
ce
s
s
with
L
AS
SO
an
d
R
F
E
h
ig
h
lig
h
ted
cr
itical
f
ea
tu
r
es
r
elev
an
t
to
NSC
L
C
s
tag
e
class
if
icatio
n
,
as sh
o
wn
in
T
ab
les
4
an
d
5
,
wh
ic
h
d
etai
l n
o
v
el
f
ea
tu
r
es id
en
tifie
d
b
y
e
ac
h
m
eth
o
d
.
T
ab
le
4
.
Selecte
d
f
ea
tu
r
es f
o
r
ea
ch
s
tag
e
b
y
L
ASSO a
lg
o
r
ith
m
S
t
a
g
e
1
S
t
a
g
e
2
S
t
a
g
e
3
F
e
a
t
u
r
e
s
F
l
a
t
n
e
ss
,
M
a
j
o
r
A
x
i
s
Le
n
g
t
h
,
e
n
e
r
g
y
,
i
n
t
e
r
q
u
a
r
t
i
l
e
r
a
n
g
e
,
mea
n
,
c
o
a
r
se
n
e
ss
,
st
r
e
n
g
t
h
9
0
P
e
r
c
e
n
t
i
l
e
,
M
a
x
i
m
u
m,
I
mc1
,
I
mc2
,
S
i
z
e
Z
o
n
e
N
o
n
-
U
n
i
f
o
r
m
i
t
y
F
l
a
t
n
e
ss
,
m
a
j
o
r
a
x
i
s
l
e
n
g
t
h
,
e
n
e
r
g
y
,
med
i
a
n
,
r
o
o
t
m
e
a
n
sq
u
a
r
e
d
,
S
k
e
w
n
e
ss
,
I
mc1
,
G
r
a
y
Le
v
e
l
V
a
r
i
a
n
c
e
.
2
,
l
o
w
g
r
a
y
l
e
v
e
l
z
o
n
e
e
mp
h
a
s
i
s,
z
o
n
e
e
n
t
r
o
p
y
,
c
o
a
r
s
e
n
e
ss,
s
t
r
e
n
g
t
h
T
ab
le
5
.
Selecte
d
f
ea
tu
r
es f
o
r
ea
ch
s
tag
e
b
y
R
FE
alg
o
r
ith
m
S
t
a
g
e
1
S
t
a
g
e
2
S
t
a
g
e
3
F
e
a
t
u
r
e
s
M
a
j
o
r
A
x
i
sL
e
n
g
t
h
,
M
a
x
i
m
u
m
2
D
D
i
a
met
e
r
R
o
w
,
M
a
x
i
m
u
m
3
D
D
i
a
met
e
r
,
S
p
h
e
r
i
c
i
t
y
,
1
0
P
e
r
c
e
n
t
i
l
e
,
R
o
o
t
M
e
a
n
S
q
u
a
r
e
d
,
C
l
u
s
t
e
r
P
r
o
mi
n
e
n
c
e
,
C
o
r
r
e
l
a
t
i
o
n
,
I
mc1
,
Jo
i
n
t
E
n
t
r
o
p
y
,
S
mal
l
D
e
p
e
n
d
e
n
c
e
L
o
w
G
r
a
y
Le
v
e
l
Em
p
h
a
s
i
s
,
R
u
n
En
t
r
o
p
y
,
Lo
w
G
r
a
y
Le
v
e
l
Z
o
n
e
Em
p
h
a
si
s
,
S
i
z
e
Z
o
n
e
N
o
n
U
n
i
f
o
r
mi
t
y
N
o
r
ma
l
i
z
e
d
,
S
mal
l
A
r
e
a
H
i
g
h
G
r
a
y
Le
v
e
l
Em
p
h
a
si
s
M
a
j
o
r
A
x
i
sL
e
n
g
t
h
,
S
p
h
e
r
i
c
i
t
y
,
9
0
P
e
r
c
e
n
t
i
l
e
,
C
l
u
s
t
e
r
P
r
o
m
i
n
e
n
c
e
,
C
l
u
st
e
r
S
h
a
d
e
,
C
l
u
st
e
r
Te
n
d
e
n
c
y
,
M
C
C
,
S
u
mA
v
e
r
a
g
e
,
G
r
a
y
Le
v
e
l
V
a
r
i
a
n
c
e
,
H
i
g
h
G
r
a
y
Le
v
e
l
Em
p
h
a
si
s
,
S
mal
l
D
e
p
e
n
d
e
n
c
e
H
i
g
h
G
r
a
y
Le
v
e
l
Emp
h
a
s
i
s,
G
r
a
y
L
e
v
e
l
V
a
r
i
a
n
c
e
.
2
,
Lo
w
G
r
a
y
Le
v
e
l
Z
o
n
e
Em
p
h
a
si
s
,
Zo
n
e
E
n
t
r
o
p
y
,
C
o
m
p
l
e
x
i
t
y
F
l
a
t
n
e
ss
,
L
e
a
s
t
A
x
i
sLe
n
g
t
h
,
M
a
j
o
r
A
x
i
sL
e
n
g
t
h
,
M
e
a
n
,
R
o
o
t
M
e
a
n
S
q
u
a
r
e
d
,
U
n
i
f
o
r
m
i
t
y
,
I
mc1
,
M
C
C
,
D
e
p
e
n
d
e
n
c
e
E
n
t
r
o
p
y
,
D
e
p
e
n
d
e
n
c
e
V
a
r
i
a
n
c
e
,
G
r
a
y
Le
v
e
l
N
o
n
U
n
i
f
o
r
mi
t
y
N
o
r
mal
i
z
ed
G
r
a
y
Le
v
e
l
N
o
n
U
n
i
f
o
r
mi
t
y
N
o
r
mal
i
z
e
d
.
1
,
G
r
a
y
Le
v
e
l
V
a
r
i
a
n
c
e
.
2
,
H
i
g
h
G
r
a
y
Le
v
e
l
Z
o
n
e
Em
p
h
a
si
s
,
S
mal
l
A
r
e
a
L
o
w
G
r
a
y
Le
v
e
l
Em
p
h
a
si
s
T
ab
les
6
an
d
7
s
h
o
ws
th
e
r
es
u
lts
o
f
th
e
f
iv
e
m
o
d
els
f
o
r
s
tag
in
g
o
f
NSC
L
C
p
atien
ts
u
s
in
g
L
ASSO
an
d
R
FE
f
ea
tu
r
e
s
elec
tio
n
m
e
th
o
d
s
.
T
h
e
ar
ea
u
n
d
er
th
e
c
u
r
v
e
(
AUC)
o
f
th
e
L
R
,
SVM,
DT
,
R
F
,
an
d
XGB
m
o
d
els
u
s
in
g
L
ASSO
was
0
.
7
7
3
9
,
0
.
7
5
7
8
,
0
.
6
8
8
6
,
0
.
8
4
2
8
,
an
d
0
.
8
1
7
0
.
T
h
e
AUC
o
f
th
e
L
R
,
SVM,
DT
,
R
F
,
an
d
XGB m
o
d
els u
s
in
g
R
FE
was 0
.
8
3
8
5
,
0
.
8
1
1
9
,
0
.
7
9
9
4
,
0
.
9
3
0
7
,
an
d
0
.
8
1
0
0
.
T
h
e
R
F
class
if
ier
,
co
m
b
in
ed
with
R
FE
f
ea
tu
r
e
s
elec
tio
n
,
d
em
o
n
s
tr
ated
th
e
b
est
p
er
f
o
r
m
a
n
ce
am
o
n
g
th
e
class
if
ier
s
in
o
u
r
s
tu
d
y
.
T
h
e
R
F
m
o
d
el
f
o
r
p
r
ed
ictin
g
t
h
e
s
tag
es
o
f
NSC
L
C
p
atien
ts
h
ad
an
ac
cu
r
ac
y
o
f
0
.
8
1
1
4
,
s
en
s
itiv
ity
o
f
0
.
8
1
3
5
,
s
p
ec
if
icity
o
f
0
.
8
1
5
0
,
F1
-
s
co
r
e
o
f
0
.
8
1
4
2
an
d
an
AUC o
f
0
.
9
3
0
7
.
T
h
e
co
n
f
u
s
io
n
m
atr
ices
an
d
r
ec
eiv
er
o
p
e
r
at
in
g
ch
ar
ac
ter
is
tic
(
R
OC
)
p
lo
ts
ar
e
s
h
o
wn
in
Fig
u
r
e
s
3
to
7
d
em
o
n
s
tr
ate
a
co
m
p
ar
is
o
n
o
f
h
o
w
well
ea
ch
class
if
icatio
n
m
eth
o
d
p
er
f
o
r
m
ed
wh
en
u
s
in
g
f
ea
tu
r
e
s
elec
tio
n
with
L
ASSO;
th
is
s
h
o
ws th
e
en
s
em
b
le
m
o
d
e
ls
p
er
f
o
r
m
e
d
b
etter
th
a
n
th
e
li
n
ea
r
m
o
d
els at
class
if
y
in
g
NSC
L
C
s
tag
es.
I
n
a
d
d
itio
n
,
th
e
co
n
f
u
s
io
n
m
atr
ices
an
d
R
OC
p
lo
ts
o
b
tain
ed
f
r
o
m
f
ea
tu
r
e
s
elec
tio
n
wit
h
R
FE
ar
e
d
em
o
n
s
tr
ated
in
Fig
u
r
es
8
to
1
2
.
T
h
e
r
esu
lts
s
h
o
wn
in
th
ese
p
lo
ts
in
d
icate
th
at
th
e
R
F
cl
ass
if
ier
was
ab
le
to
class
if
y
all
o
f
th
e
d
if
f
er
e
n
t stag
es with
th
e
b
est d
is
cr
im
in
atio
n
.
T
h
is
is
r
ef
lecte
d
b
y
its
h
ig
h
est AUC.
T
ab
le
6
.
R
esu
lts
o
f
th
e
m
o
d
els f
o
r
s
tag
in
g
o
f
NSC
L
C
p
atien
ts
with
L
ASSO
C
l
a
s
si
f
i
e
r
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
Te
st
a
c
c
u
r
a
c
y
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
S
e
n
s
i
t
i
v
i
t
y
S
p
e
c
i
f
i
c
i
t
y
F1
-
s
c
o
r
e
AUC
LR
0
.
5
9
5
0
0
.
5
4
9
1
0
.
5
8
9
0
0
.
5
4
9
9
0
.
5
4
7
6
0
.
5
4
8
8
0
.
7
7
3
9
S
V
M
0
.
5
7
6
4
0
.
5
5
7
3
0
.
5
7
2
5
0
.
5
5
6
7
0
.
5
6
4
5
0
.
5
6
0
6
0
.
7
5
7
8
DT
1
.
0
0
.
5
8
1
9
0
.
5
4
9
6
0
.
5
8
5
9
0
.
5
8
6
1
0
.
5
8
6
0
0
.
6
8
8
6
RF
0
.
8
8
0
1
0
.
6
8
0
3
0
.
6
3
8
8
0
.
6
8
6
6
0
.
6
9
0
3
0
.
6
8
8
4
0
.
8
4
2
8
XGB
1
.
0
0
.
6
3
9
3
0
.
6
9
7
2
0
.
6
4
6
1
0
.
6
4
2
8
0
.
6
4
4
4
0
.
8
1
7
0
T
ab
le
7
.
R
esu
lts
o
f
th
e
m
o
d
els f
o
r
s
tag
in
g
o
f
NSC
L
C
p
atien
ts
with
R
FE
C
l
a
s
si
f
i
e
r
Tr
a
i
n
i
n
g
a
c
c
u
r
a
c
y
Te
st
a
c
c
u
r
a
c
y
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
S
e
n
s
i
t
i
v
i
t
y
S
p
e
c
i
f
i
c
i
t
y
F
1
S
c
o
r
e
AUC
LR
0
.
7
2
3
1
0
.
6
8
0
3
0
.
6
9
6
0
0
.
6
7
9
8
0
.
6
9
2
9
0
.
6
8
6
3
0
.
8
3
8
5
S
V
M
0
.
6
0
5
3
0
.
5
8
1
9
0
.
5
9
7
2
0
.
5
7
4
6
0
.
5
9
6
5
0
.
5
8
5
3
0
.
8
1
1
9
DT
1
.
0
0
.
7
2
9
5
0
.
7
0
2
8
0
.
7
3
3
7
0
.
7
3
2
8
0
.
7
3
3
3
0
.
7
9
9
4
RF
1
.
0
0
.
8
1
1
4
0
.
7
7
0
6
0
.
8
1
3
5
0
.
8
1
5
0
0
.
8
1
4
2
0
.
9
3
0
7
XGB
1
.
0
0
.
6
0
6
5
0
.
6
6
1
0
0
.
6
0
8
8
0
.
6
0
1
9
0
.
6
0
5
3
0
.
8
1
0
0
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
C
o
mp
u
ted
to
mo
g
r
a
p
h
y
ima
g
in
g
r
a
d
io
mics:
a
n
o
ve
l a
p
p
r
o
a
c
h
to
ea
r
ly
-
s
ta
g
e
n
o
n
-
s
ma
ll …
(
R
a
viteja
B
a
leka
i
)
2477
Fig
u
r
e
3
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
L
R
u
s
in
g
L
A
SS
O
in
NS
C
L
C
s
tag
e
s
class
if
i
ca
tio
n
Fig
u
r
e
4
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
SVM
u
s
in
g
L
ASSO in
NS
C
L
C
s
tag
es c
las
s
i
f
icatio
n
Fig
u
r
e
5
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
DT
u
s
in
g
L
A
SS
O
in
NS
C
L
C
s
tag
e
s
class
if
i
ca
tio
n
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.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
4
7
1
-
2
4
8
3
2478
Fig
u
r
e
6
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
R
F u
s
in
g
L
A
SS
O
in
NS
C
L
C
s
tag
e
s
class
if
i
ca
tio
n
Fig
u
r
e
7
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
XGBo
o
s
t u
s
in
g
L
ASSO in
NSC
L
C
s
tag
es
class
if
icatio
n
Fig
u
r
e
8
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
L
R
u
s
in
g
R
FE
in
NSC
L
C
s
tag
e
s
class
if
icat
io
n
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
C
o
mp
u
ted
to
mo
g
r
a
p
h
y
ima
g
in
g
r
a
d
io
mics:
a
n
o
ve
l a
p
p
r
o
a
c
h
to
ea
r
ly
-
s
ta
g
e
n
o
n
-
s
ma
ll …
(
R
a
viteja
B
a
leka
i
)
2479
Fig
u
r
e
9
.
C
o
n
f
u
s
io
n
m
atr
i
x
an
d
R
OC
cu
r
v
e
o
f
SVM
u
s
in
g
R
FE
in
NSC
L
C
s
tag
es c
las
s
if
ic
atio
n
Fig
u
r
e
1
0
.
C
o
n
f
u
s
io
n
m
atr
ix
a
n
d
R
OC
cu
r
v
e
o
f
DT
u
s
in
g
R
FE
in
NSC
L
C
s
tag
es c
las
s
if
ic
atio
n
Fig
u
r
e
1
1
.
C
o
n
f
u
s
io
n
m
atr
ix
a
n
d
R
OC
cu
r
v
e
o
f
R
F u
s
in
g
R
FE
in
NSC
L
C
s
tag
e
s
class
if
icat
io
n
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.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
4
7
1
-
2
4
8
3
2480
Fig
u
r
e
1
2
.
C
o
n
f
u
s
io
n
m
atr
ix
a
n
d
R
OC
cu
r
v
e
o
f
XGBo
o
s
t u
s
in
g
R
FE
in
NSC
L
C
s
tag
es c
la
s
s
if
icatio
n
3
.
2
.
Dis
cus
s
io
n
I
n
r
ec
en
t
y
ea
r
s
,
r
ad
io
m
ics
-
b
a
s
ed
m
o
d
els
o
f
NSC
L
C
h
is
to
l
o
g
ical
an
d
clin
ical
s
tag
in
g
class
if
icatio
n
h
av
e
o
b
tain
ed
co
n
s
id
er
ab
le
i
n
ter
est
o
win
g
to
th
eir
ca
p
ac
i
ty
to
q
u
a
n
titativ
ely
an
aly
ze
t
u
m
o
r
p
r
o
p
er
ties
in
ter
m
s
o
f
im
ag
in
g
i
n
f
o
r
m
atio
n
.
T
h
e
m
o
d
els
ar
e
a
n
o
n
-
in
v
asiv
e,
ef
f
icien
t,
an
d
o
b
jectiv
e
alter
n
ativ
e
to
th
e
co
n
v
en
tio
n
al
s
tag
in
g
p
r
o
ce
d
u
r
es
lik
e
b
io
p
s
y
an
d
m
an
u
al
im
ag
in
g
in
ter
p
r
etatio
n
.
I
n
th
at
o
u
r
m
o
d
el
is
b
u
ilt
o
n
r
ad
io
m
ic
f
ea
tu
r
es
o
b
tain
e
d
b
y
C
T
im
ag
es,
wh
ich
ca
n
p
r
ed
ic
t
th
e
s
tag
e
o
f
NSC
L
C
with
h
ig
h
ac
cu
r
a
cy
an
d
is
co
n
s
is
ten
t
with
th
e
in
cr
ea
s
in
g
liter
atu
r
e
o
n
ML
-
b
ased
ca
n
c
er
d
iag
n
o
s
tics
.
R
ec
en
t
in
v
esti
g
atio
n
s
h
av
e
s
h
o
wn
th
at
r
ad
io
m
ics
b
ased
o
n
C
T
a
n
d
ML
ca
n
d
eliv
e
r
s
tr
o
n
g
p
er
f
o
r
m
an
ce
in
ter
m
s
o
f
NSC
L
C
s
tag
in
g
an
d
r
elate
d
clin
ical
p
r
ed
ictio
n
ac
tiv
ities
[
2
4
]
,
[
2
5
]
.
So
m
e
s
tu
d
ies
h
av
e
also
p
o
in
ted
o
u
t
th
e
p
o
s
s
ib
ilit
ies
o
f
lu
n
g
ca
n
ce
r
s
tag
in
g
u
s
in
g
r
ad
io
m
ics.
Yu
et
a
l.
[
2
6
]
p
r
o
p
o
s
ed
th
e
m
o
d
el
to
p
r
e
d
ict
th
e
p
ath
o
l
o
g
ic
al
s
tag
e
(
I
A
g
r
ad
e
to
I
V)
o
f
N
SC
L
C
b
a
s
ed
o
n
C
T
r
ad
io
m
ic
f
ea
tu
r
es,
an
d
r
ep
o
r
ted
th
at
a
R
F
cla
s
s
if
ier
h
ad
b
etter
av
er
ag
e
p
r
ec
is
io
n
in
b
o
th
lu
n
g
ADC
(
AP
=0
.
8
4
)
a
n
d
lu
n
g
SC
C
(
AP
=0
.
6
2
)
.
Nev
e
r
th
eless
,
th
eir
r
esear
ch
in
d
icate
d
th
at
t
h
ey
h
ad
lim
itatio
n
s
ass
o
ciate
d
with
th
e
g
en
er
aliza
b
ilit
y
o
f
th
e
m
o
d
el
an
d
o
v
er
f
itti
n
g
.
T
h
ese
f
in
d
i
n
g
s
h
av
e
b
e
en
s
tr
en
g
t
h
en
ed
b
y
m
o
r
e
r
ec
e
n
t
m
u
lticen
ter
s
tu
d
i
es
th
at
h
av
e
also
u
n
d
er
s
co
r
e
d
th
e
n
ee
d
to
h
av
e
s
tr
o
n
g
v
alid
a
tio
n
p
lan
s
th
at
ca
n
g
u
ar
an
tee
th
e
g
en
e
r
aliza
b
ilit
y
o
f
r
esu
lts
to
h
eter
o
g
en
eo
u
s
d
at
asets
[
2
7
]
,
[
2
8
]
.
S
h
i
m
a
d
a
e
t
a
l.
[
2
9
]
i
n
v
e
s
t
i
g
a
t
e
d
th
e
u
s
e
o
f
C
T
-
b
a
s
ed
r
ad
i
o
m
ic
s
w
i
t
h
an
AI
-
b
a
s
e
d
ap
p
r
o
a
ch
to
p
r
ed
i
c
t
t
h
e
p
a
th
o
lo
g
i
ca
l
l
y
m
p
h
n
o
d
e
m
e
t
a
s
t
a
s
i
s
i
n
e
ar
l
y
-
s
t
ag
e
N
S
C
L
C
w
i
t
h
a
n
ac
c
u
r
a
cy
o
f
0
.
6
5
.
S
i
m
i
la
r
l
y
,
Y
i
n
e
t
a
l
.
[
3
0
]
d
e
m
o
n
s
t
r
a
t
e
d
t
h
a
t
th
e
u
s
e
o
f
C
T
r
a
d
io
m
i
c
s
c
o
m
b
i
n
ed
w
i
th
d
e
ep
l
e
a
r
n
i
n
g
a
p
p
r
o
a
ch
e
s
ca
n
b
e
u
s
e
d
to
p
r
e
d
i
c
t
l
y
m
p
h
n
o
d
e
m
e
t
a
s
t
a
s
i
s
o
f
e
a
r
ly
-
s
t
ag
e
lu
n
g
A
D
C
m
o
r
e
a
c
cu
r
a
t
e
ly
,
w
h
i
c
h
l
e
ad
s
t
o
t
h
e
g
r
o
w
in
g
i
m
p
o
r
t
a
n
ce
o
f
h
i
g
h
er
M
L
t
e
c
h
n
i
q
u
e
s
in
im
p
r
o
v
in
g
s
t
a
g
in
g
r
e
s
u
l
t
s
.
Ub
a
l
d
i
e
t
a
l
.
[
3
1
]
u
s
e
d
a
M
L
m
o
d
e
l
t
o
p
r
ed
i
c
t
l
u
n
g
ca
n
c
er
s
t
a
g
e
s
(
I
to
I
I
)
an
d
h
i
s
t
o
l
o
g
y
w
i
t
h
l
i
m
i
te
d
s
a
m
p
l
e
s
o
f
C
T
d
a
t
a
w
i
th
A
U
C
v
a
l
u
e
s
o
f
0
.
7
2
±
0
.
0
4
w
i
t
h
th
e
R
F
a
n
d
0
.
8
4
±
0
.
0
3
w
i
th
a
l
i
n
e
ar
S
V
M
.
L
i
n
e
t
a
l
.
[
3
2
]
c
am
e
u
p
w
i
th
a
m
o
d
e
l
t
h
a
t
c
a
te
g
o
r
i
z
e
s
b
o
th
h
i
s
t
o
l
o
g
i
c
a
l
s
u
b
ty
p
e
a
n
d
c
l
i
n
i
c
a
l
s
t
a
g
e
(
I
t
o
I
I
I
)
o
f
N
S
C
L
C
w
i
t
h
a
s
t
a
g
in
g
A
U
C
o
f
0
.
8
8
1
.
R
e
c
en
t
s
t
u
d
ies
in
[
2
4
]
,
[
3
3
]
h
a
v
e
a
l
s
o
s
h
o
w
n
t
h
a
t
C
T
r
a
d
io
m
i
c
s
-
b
a
s
e
d
M
L
c
l
a
s
s
i
f
i
e
r
s
,
in
p
a
r
t
i
cu
l
a
r
t
h
e
RF
c
l
a
s
s
i
f
i
e
r
s
,
c
o
u
ld
a
l
s
o
b
e
s
t
r
o
n
g
an
d
s
ta
b
l
e
in
t
h
e
s
t
a
g
in
g
o
f
N
S
C
L
C
w
h
e
n
te
s
t
e
d
o
n
m
u
l
t
i
c
en
t
e
r
d
a
ta
s
e
t
s
.
T
h
e
m
o
d
el
will
in
c
o
r
p
o
r
ate
s
o
m
e
o
f
th
e
lim
itatio
n
s
th
at
wer
e
o
b
s
er
v
e
d
in
ea
r
lier
r
esear
ch
es.
F
ea
tu
r
e
s
elec
tio
n
with
L
AS
SO
an
d
R
FE
was
u
s
ed
a
s
th
e
in
itial
s
ta
g
e
o
f
d
ata
d
im
en
s
io
n
ality
r
ed
u
ctio
n
,
an
d
th
e
latter
is
im
p
o
r
tan
t
to
in
cr
ea
s
e
th
e
lik
elih
o
o
d
o
f
i
n
ter
p
r
etin
g
th
e
r
esu
lts
s
in
ce
R
FE
o
n
ly
co
n
s
id
er
s
th
e
m
o
s
t
s
ig
n
if
ican
t
r
ad
io
m
ic
f
ea
tu
r
es
t
o
s
tag
e
th
e
d
ata.
T
h
e
s
u
cc
ess
o
f
L
ASSO
-
b
ased
f
ea
t
u
r
e
s
el
ec
tio
n
in
r
a
d
io
m
ics
h
as
b
ee
n
ad
e
q
u
ately
r
ep
o
r
ted
.
Pas
in
i
et
a
l.
[
3
4
]
d
e
m
o
n
s
tr
ated
th
at
L
ASSO
was
ef
f
ec
tiv
e
in
elim
in
atin
g
ir
r
elev
an
t
f
ea
tu
r
es
an
d
e
n
h
an
cin
g
m
o
d
el
ac
cu
r
ac
y
to
p
r
ed
i
ct
h
is
to
p
ath
o
lo
g
ical
s
u
b
t
y
p
es
o
f
NSC
L
C
.
On
th
e
s
am
e
n
o
te,
th
e
r
esu
lts
in
d
icate
th
at
th
e
ad
d
itio
n
o
f
L
ASSO o
r
R
FE
w
ith
en
s
em
b
le
-
b
ased
class
if
ier
s
,
e
s
p
ec
ially
th
e
R
F,
r
esu
lts
in
g
r
ea
ter
s
tag
in
g
p
er
f
o
r
m
a
n
ce
.
T
h
e
s
ec
o
n
d
im
p
o
r
ta
n
t
im
p
r
o
v
em
en
t
o
f
o
u
r
r
esear
ch
is
th
e
d
ir
ec
t
m
an
ag
em
en
t
o
f
th
e
class
im
b
alan
ce
th
at
is
an
in
h
er
en
t
p
r
o
b
lem
in
m
ed
ical
im
ag
in
g
d
ata
in
w
h
ich
s
p
ec
if
ic
ca
n
ce
r
s
tag
es
ar
e
u
n
d
er
r
e
p
r
esen
ted
.
Mo
d
els
th
at
a
r
e
tr
ai
n
ed
with
u
n
b
alan
ce
d
s
am
p
les
will
g
i
v
e
a
d
v
an
tag
e
to
m
ajo
r
ity
class
es
an
d
e
v
en
tu
ally
m
ak
e
b
iased
p
r
e
d
ictio
n
s
as
o
th
er
p
r
ev
io
u
s
r
a
d
io
m
ics
s
tu
d
ies
h
av
e
o
b
s
er
v
e
d
[
3
5
]
.
T
h
is
p
r
o
b
lem
is
s
o
lv
ed
b
y
u
s
in
g
th
e
SMOT
E
to
tr
ain
o
n
ly
th
e
tr
ain
in
g
d
ata,
an
d
th
is
s
o
lv
e
d
th
e
p
r
o
b
lem
a
n
d
th
e
class
if
icatio
n
was
b
etter
,
Evaluation Warning : The document was created with Spire.PDF for Python.