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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
,
p
p
.
5
1
0
6
~
5
1
1
9
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
6
.
p
p
5
1
0
6
-
5
1
1
9
5106
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
O
ptimi
zing
brain
tumo
r MRI cla
ss
i
ficatio
n using
adv
a
nced
prepro
cess
ing
t
e
c
hniques
and
ense
mble lea
rning
m
e
thods
Ak
im
M
a
na
o
r
H
a
ra
P
a
rdede
1
,
Ahm
a
d Z
a
m
s
uri
2
,
I
nd
i N
uro
ini
3
,
P
utr
a
m
a
Alk
ha
iri
4
1
I
n
f
o
r
mat
i
o
n
S
y
st
e
ms St
u
d
y
P
r
o
g
r
a
m
,
S
TM
I
K
K
a
p
u
t
a
ma
,
B
i
n
j
a
i
,
I
n
d
o
n
e
si
a
2
M
a
s
t
e
r
P
r
o
g
r
a
m
of
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
a
s
La
n
c
a
n
g
K
u
n
i
n
g
,
P
e
k
a
n
b
a
r
u
,
I
n
d
o
n
e
si
a
3
F
a
c
u
l
t
y
o
f
La
w
,
U
n
i
v
e
r
si
t
a
s
B
h
a
y
a
n
g
k
a
r
a
S
u
r
a
b
a
y
a
,
S
u
r
a
b
a
y
a
,
I
n
d
o
n
e
si
a
4
I
n
f
o
r
mat
i
c
s
M
a
n
a
g
e
me
n
t
,
S
TI
K
O
M
Tu
n
a
s
B
a
n
g
sa,
P
e
m
a
t
a
n
g
si
a
n
t
a
r
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
1
1
,
2
0
2
5
R
ev
is
ed
Oct
2
7
,
2
0
2
5
Acc
ep
ted
No
v
8
,
2
0
2
5
Bra
in
t
u
m
o
r
c
las
sifica
ti
o
n
is
a
c
r
it
ica
l
tas
k
in
m
e
d
ica
l
ima
g
i
n
g
th
a
t
d
irec
tl
y
imp
a
c
ts
th
e
a
c
c
u
ra
c
y
o
f
d
iag
n
o
sis
a
n
d
trea
tme
n
t
p
lan
n
in
g
.
H
o
we
v
e
r,
th
e
c
o
m
p
lex
it
y
a
n
d
v
a
riab
i
li
ty
o
f
m
a
g
n
e
ti
c
re
so
n
a
n
c
e
ima
g
i
n
g
(
M
R
I
)
ima
g
e
s
p
o
se
si
g
n
if
ica
n
t
c
h
a
ll
e
n
g
e
s,
o
fte
n
re
su
lt
i
n
g
i
n
re
d
u
c
e
d
m
o
d
e
l
re
li
a
b
il
it
y
a
n
d
g
e
n
e
ra
li
z
a
ti
o
n
.
T
h
is
stu
d
y
a
d
d
re
s
se
s
th
e
se
li
m
it
a
ti
o
n
s
b
y
p
r
o
p
o
sin
g
a
n
o
v
e
l
Re
sN
e
t+Ba
g
g
in
g
m
o
d
e
l,
lev
e
ra
g
i
n
g
t
h
e
stre
n
g
th
s
o
f
re
sid
u
a
l
n
e
t
wo
rk
s
a
n
d
e
n
se
m
b
le
lea
rn
in
g
t
o
e
n
h
a
n
c
e
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
.
Us
in
g
p
u
b
li
c
ly
a
v
a
il
a
b
le
b
ra
in
tu
m
o
r
M
RI
d
a
ta
se
ts,
in
c
lu
d
i
n
g
ima
g
e
s
lab
e
led
a
s
b
e
n
ig
n
,
m
a
li
g
n
a
n
t,
a
n
d
n
o
rm
a
l,
t
h
e
stu
d
y
e
m
p
l
o
y
s
a
d
v
a
n
c
e
d
p
re
p
ro
c
e
ss
in
g
tec
h
n
iq
u
e
s
su
c
h
a
s
n
o
rm
a
li
z
a
ti
o
n
,
d
a
ta
a
u
g
m
e
n
tatio
n
,
a
n
d
n
o
ise
re
d
u
c
ti
o
n
to
e
n
su
re
h
ig
h
-
q
u
a
li
t
y
in
p
u
ts.
Th
e
p
ro
p
o
se
d
m
o
d
e
l
d
e
m
o
n
stra
ted
sig
n
ifi
c
a
n
t
imp
ro
v
e
m
e
n
ts,
a
c
h
ie
v
in
g
t
h
e
h
i
g
h
e
st t
e
stin
g
a
c
c
u
ra
c
y
o
f
7
2
%
,
o
u
tp
e
rfo
rm
in
g
o
th
e
r
tes
ted
m
o
d
e
ls
su
c
h
a
s
LeN
e
t,
sta
n
d
a
rd
Re
sN
e
t,
G
o
o
g
leN
e
t,
a
n
d
VG
GN
e
t.
P
re
c
isio
n
(0
.
6
0
1
0
),
re
c
a
ll
(0
.
6
0
0
0
),
a
n
d
F
1
-
s
c
o
re
(0
.
5
9
9
0
)
m
e
tri
c
s
fu
rth
e
r
h
i
g
h
l
ig
h
t
it
s
s
u
p
e
rio
r
b
a
lan
c
e
in
d
e
tec
ti
n
g
p
o
sit
iv
e
a
n
d
n
e
g
a
ti
v
e
c
las
se
s.
Th
e
n
o
v
e
lt
y
o
f
t
h
is
re
se
a
rc
h
li
e
s
i
n
t
h
e
a
p
p
li
c
a
ti
o
n
o
f
B
a
g
g
in
g
to
Re
sN
e
t,
wh
ich
e
ffe
c
ti
v
e
ly
m
it
i
g
a
tes
o
v
e
rfit
ti
n
g
a
n
d
e
n
h
a
n
c
e
s
p
re
d
icti
v
e
sta
b
il
it
y
i
n
c
o
m
p
lex
m
e
d
ica
l
d
a
tas
e
ts.
Th
e
se
fi
n
d
i
n
g
s
u
n
d
e
r
sc
o
re
th
e
p
ro
p
o
se
d
m
o
d
e
l'
s p
o
ten
ti
a
l
a
s a
ro
b
u
st so
lu
ti
o
n
fo
r
b
ra
in
t
u
m
o
r
c
las
sifica
ti
o
n
,
c
o
n
tri
b
u
ti
n
g
to
m
o
re
a
c
c
u
ra
te an
d
re
li
a
b
le d
iag
n
o
st
ics
.
K
ey
w
o
r
d
s
:
B
ag
g
in
g
B
r
ain
tu
m
o
r
class
if
icatio
n
E
n
s
em
b
le
lear
n
in
g
Me
d
ical
im
ag
e
an
aly
s
is
MRI
im
ag
es
R
esNet
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
:
Ak
im
Ma
n
ao
r
Har
a
Par
d
ed
e
I
n
f
o
r
m
atio
n
Sy
s
tem
s
Stu
d
y
Pr
o
g
r
am
,
STM
I
K
Kap
u
tam
a
St
.
Vete
r
an
No
.
4
A
-
9
A,
B
in
jai
-
No
r
th
Su
m
ater
a,
I
n
d
o
n
esia
E
m
ail: a
k
im
m
h
p
@
liv
e.
co
m
1.
I
NT
RO
D
UCT
I
O
N
B
r
ain
tu
m
o
r
s
ar
e
ab
n
o
r
m
al
c
ell
g
r
o
wth
s
with
in
o
r
a
r
o
u
n
d
b
r
ain
tis
s
u
e
th
at
ca
n
b
e
eith
er
b
en
i
g
n
(
n
o
n
-
c
an
ce
r
o
u
s
)
o
r
m
alig
n
an
t
(
ca
n
ce
r
o
u
s
)
[
1
]
,
[
2
]
.
T
h
is
c
o
n
d
itio
n
c
an
i
n
ter
f
er
e
with
v
ital
b
r
ain
f
u
n
ctio
n
s
,
s
u
ch
as
co
g
n
itiv
e
ab
ilit
ies,
m
o
to
r
co
o
r
d
in
atio
n
,
an
d
s
en
s
o
r
y
s
y
s
tem
s
,
d
ep
en
d
in
g
o
n
t
h
e
t
u
m
o
r
'
s
lo
ca
tio
n
an
d
s
ize.
B
r
ain
tu
m
o
r
s
ca
n
d
ev
elo
p
p
r
im
ar
ily
in
th
e
b
r
ai
n
o
r
as
m
etastas
es
f
r
o
m
ca
n
ce
r
in
o
th
er
p
ar
ts
o
f
th
e
b
o
d
y
.
Sy
m
p
to
m
s
v
ar
y
,
r
an
g
in
g
f
r
o
m
wo
r
s
en
in
g
h
ea
d
ac
h
es,
s
eizu
r
es,
b
lu
r
r
e
d
v
is
io
n
,
to
m
e
m
o
r
y
im
p
air
m
en
ts
[
3
]
,
[
4
]
.
E
ar
ly
d
iag
n
o
s
is
is
cr
u
cial
f
o
r
im
p
r
o
v
i
n
g
th
e
ch
an
ce
s
o
f
ef
f
ec
tiv
e
tr
ea
tm
en
t,
an
d
m
ed
ical
im
ag
in
g
s
u
ch
as
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
(
MRI
)
is
a
k
ey
to
o
l
f
o
r
d
etec
tin
g
an
d
ev
al
u
atin
g
b
r
a
in
tu
m
o
r
s
.
W
ith
th
e
r
ig
h
t
ap
p
r
o
ac
h
,
in
cl
u
d
in
g
m
ed
ical
in
ter
v
en
tio
n
s
an
d
th
e
u
s
e
o
f
ar
tific
ial
in
tellig
e
n
ce
(
AI
)
f
o
r
im
ag
e
an
aly
s
is
,
p
atien
t
r
ec
o
v
er
y
ch
an
ce
s
ca
n
s
ig
n
if
ican
tly
im
p
r
o
v
e
.
B
r
ain
ca
n
ce
r
is
o
n
e
o
f
th
e
d
is
ea
s
es
w
ith
a
h
ig
h
m
o
r
tality
r
ate,
esp
ec
ially
if
n
o
t
d
etec
t
ed
an
d
tr
ea
ted
at
an
ea
r
ly
s
tag
e.
E
ar
ly
d
etec
tio
n
o
f
b
r
ai
n
tu
m
o
r
s
p
lay
s
an
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
Op
timiz
in
g
b
r
a
in
tu
mo
r
MRI
cla
s
s
ifica
tio
n
u
s
in
g
a
d
va
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
… (
A
kim
Ma
n
a
o
r
Ha
r
a
P
a
r
d
ed
e
)
5107
im
p
o
r
tan
t
r
o
le
in
in
cr
ea
s
in
g
p
atien
ts
'
ch
an
ce
s
o
f
r
ec
o
v
er
y
.
I
n
m
ed
ical
p
r
ac
tice,
MRI
i
s
o
n
e
o
f
th
e
m
o
s
t
r
eliab
le
im
ag
in
g
tech
n
iq
u
es
f
o
r
d
etec
tin
g
an
d
a
n
aly
zin
g
b
r
ain
tu
m
o
r
s
[
5
]
–
[
7
]
.
Ho
we
v
er
,
m
an
u
al
an
aly
s
is
o
f
MRI
im
ag
es
is
o
f
te
n
tim
e
-
c
o
n
s
u
m
in
g
,
r
eq
u
ir
es
h
ig
h
ex
p
er
tis
e,
an
d
is
p
r
o
n
e
to
h
u
m
an
er
r
o
r
.
T
h
e
r
ef
o
r
e,
m
o
r
e
ef
f
ec
tiv
e
an
d
ef
f
icien
t a
p
p
r
o
ac
h
es a
r
e
n
ee
d
e
d
to
im
p
r
o
v
e
t
h
e
ac
cu
r
ac
y
o
f
b
r
ai
n
tu
m
o
r
class
if
icatio
n
[
8
]
,
[
9
]
.
I
n
b
r
ain
tu
m
o
r
class
if
icatio
n
u
s
in
g
im
ag
i
n
g
tech
n
iq
u
es
s
u
ch
as
MRI,
s
ev
er
al
d
ee
p
lear
n
in
g
ar
ch
itectu
r
es
h
av
e
p
r
o
v
en
to
b
e
r
eliab
le
an
d
wid
ely
u
s
ed
i
n
r
esear
ch
.
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs),
s
u
ch
as
L
eNe
t
[
1
0
]
,
Alex
Net
[
1
1
]
,
VGG
-
16
[
1
2
]
,
[
1
3
]
an
d
R
esNet
[
1
4
]
,
[
1
5
]
ar
e
f
r
e
q
u
en
tl
y
em
p
lo
y
ed
d
u
e
to
th
eir
ab
ilit
y
to
ca
p
tu
r
e
s
p
atial
f
ea
tu
r
es
o
f
im
ag
es.
L
eNe
t
is
s
u
itab
le
f
o
r
s
m
all
d
atasets
d
u
e
to
its
s
im
p
licity
,
wh
ile
Alex
Net
an
d
VGG1
6
ar
e
b
etter
at
h
an
d
lin
g
m
o
r
e
co
m
p
lex
im
ag
es
d
u
e
to
th
e
ir
d
ee
p
er
n
etwo
r
k
ar
ch
itectu
r
e
[
1
6
]
,
[
1
7
]
.
R
esNet,
with
its
s
k
ip
co
n
n
ec
tio
n
s
,
is
ca
p
ab
le
o
f
ad
d
r
ess
in
g
th
e
d
eg
r
ad
atio
n
p
r
o
b
lem
in
v
er
y
d
ee
p
n
etwo
r
k
s
,
th
u
s
im
p
r
o
v
in
g
class
if
icatio
n
ac
cu
r
ac
y
.
Fu
r
th
e
r
m
o
r
e,
tr
a
n
s
f
er
lear
n
in
g
-
b
ased
ar
ch
itectu
r
es,
s
u
ch
as
I
n
ce
p
ti
o
n
V3
an
d
E
f
f
icien
tNet,
ar
e
a
ls
o
p
o
p
u
lar
b
ec
au
s
e
th
ey
all
o
w
th
e
u
s
e
o
f
p
r
e
-
tr
ain
ed
m
o
d
els
o
n
lar
g
e
d
atas
ets,
wh
ich
ca
n
th
en
b
e
f
in
e
-
tu
n
ed
f
o
r
b
r
ain
tu
m
o
r
d
atasets
.
T
h
e
co
m
b
in
atio
n
o
f
C
NNs
with
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
,
s
u
ch
as
B
ag
g
in
g
,
h
as
also
p
r
o
v
e
n
ef
f
e
ctiv
e
in
im
p
r
o
v
i
n
g
class
if
icatio
n
ac
cu
r
ac
y
b
y
c
o
m
b
in
in
g
p
r
ed
ictio
n
s
f
r
o
m
m
u
l
tip
le
m
o
d
els.
Pre
v
io
u
s
r
esear
ch
b
y
[
1
8
]
in
t
h
is
ar
ticle
h
as
b
o
t
h
s
tr
en
g
t
h
s
an
d
wea
k
n
ess
es.
T
h
e
m
ain
ad
v
an
tag
e
o
f
th
is
ar
ticle
i
s
th
e
u
s
e
o
f
ad
v
an
ce
d
d
ee
p
lear
n
in
g
m
o
d
els
s
u
ch
as
3
D
U
-
Net,
P
SP
Net,
an
d
Dee
p
L
ab
V3
+,
wh
ic
h
h
av
e
s
h
o
wn
p
r
o
m
is
in
g
r
esu
lts
in
b
r
ain
tu
m
o
r
s
eg
m
en
tatio
n
f
r
o
m
MRI
im
ag
es,
with
3
D
U
-
Net
ac
h
iev
in
g
th
e
h
ig
h
est
d
ice
s
im
ilar
ity
co
ef
f
ic
ien
t
(
DSC
)
o
f
0
.
9
0
.
A
d
d
itio
n
ally
,
th
e
ar
ticle
d
is
cu
s
s
es
th
e
im
p
o
r
tan
ce
o
f
d
ata
au
g
m
en
tatio
n
an
d
tr
a
n
s
f
er
lea
r
n
in
g
tech
n
iq
u
es
in
im
p
r
o
v
in
g
m
o
d
el
ac
cu
r
ac
y
,
w
h
ich
h
av
e
p
r
o
v
e
n
ef
f
ec
tiv
e
in
en
h
an
cin
g
m
o
d
el
p
er
f
o
r
m
a
n
c
e.
Ho
wev
er
,
th
e
ar
ticle'
s
d
r
awb
ac
k
s
in
clu
d
e
th
e
u
s
e
o
f
a
s
in
g
le
d
ataset,
n
am
ely
B
r
aT
S
2
0
1
8
,
wh
ich
m
a
y
lim
i
t
th
e
g
en
er
aliza
b
ilit
y
o
f
th
e
f
in
d
in
g
s
,
a
n
d
a
lack
o
f
in
-
d
ep
th
h
y
p
er
p
a
r
am
eter
ev
alu
atio
n
.
Mo
r
eo
v
er
,
alth
o
u
g
h
3
D
U
-
Net
d
em
o
n
s
tr
ated
t
h
e
b
est
p
er
f
o
r
m
an
ce
,
th
is
m
o
d
el
also
h
as
h
ig
h
co
m
p
u
tatio
n
al
r
eq
u
i
r
em
en
ts
a
n
d
lo
n
g
er
tr
ai
n
in
g
tim
es
c
o
m
p
ar
ed
to
o
th
er
m
o
d
els
lik
e
R
esNet5
0
,
wh
ich
is
ea
s
ier
to
im
p
lem
en
t.
On
th
e
o
t
h
er
h
an
d
,
th
e
s
tu
d
y
b
y
[
1
9
]
h
a
s
b
o
th
s
ig
n
if
ican
t
s
tr
en
g
th
s
an
d
wea
k
n
ess
es.
T
h
e
m
ain
ad
v
an
ta
g
e
o
f
t
h
is
s
tu
d
y
is
th
e
u
s
e
o
f
tr
an
s
f
er
lear
n
in
g
m
eth
o
d
s
th
at
h
av
e
p
r
o
v
en
to
i
m
p
r
o
v
e
b
r
ain
t
u
m
o
r
class
if
icatio
n
ac
cu
r
ac
y
,
with
th
e
VGG
-
1
6
m
o
d
el
ac
h
iev
in
g
th
e
h
ig
h
est
ac
cu
r
ac
y
o
f
9
7
%
an
d
th
e
s
h
o
r
test
p
r
o
ce
s
s
in
g
tim
e
am
o
n
g
t
h
e
m
o
d
els
test
ed
,
at
2
2
%
o
f
th
e
to
tal
tim
e.
Ad
d
itio
n
ally
,
th
e
s
tu
d
y
p
er
f
o
r
m
s
a
co
m
p
r
eh
e
n
s
iv
e
co
m
p
ar
ativ
e
an
aly
s
is
o
f
tr
an
s
f
er
lea
r
n
i
n
g
m
o
d
els
s
u
ch
as
VGG
-
1
6
,
Mo
b
ileNet,
an
d
R
esNet
-
5
0
,
p
r
o
v
id
i
n
g
v
alu
ab
le
in
s
ig
h
ts
in
to
th
e
s
tr
en
g
th
s
an
d
e
f
f
icien
cies
o
f
ea
ch
m
o
d
el.
Ho
wev
e
r
,
th
e
s
tu
d
y
'
s
wea
k
n
ess
es
in
clu
d
e
th
e
lim
ited
av
ailab
ilit
y
o
f
d
iv
er
s
e
d
atasets
,
wh
ich
ca
n
a
f
f
ec
t
class
if
icatio
n
ac
cu
r
ac
y
an
d
in
tr
o
d
u
ce
b
ias,
as we
ll a
s
ch
allen
g
es in
g
en
er
al
izin
g
th
e
m
o
d
el
f
o
r
r
ar
e
t
u
m
o
r
s
u
b
ty
p
es.
I
n
r
ec
e
n
t
y
ea
r
s
,
ad
v
a
n
ce
m
en
t
s
in
AI
tech
n
o
lo
g
y
h
a
v
e
m
a
d
e
s
ig
n
if
ican
t
c
o
n
tr
ib
u
tio
n
s
t
o
m
ed
ical
im
ag
e
an
aly
s
is
[
2
0
]
,
[
2
1
]
.
S
p
ec
if
ically
,
C
NNs
h
av
e
p
r
o
v
en
th
eir
s
u
p
e
r
io
r
ity
in
p
r
o
ce
s
s
in
g
im
ag
e
d
ata,
in
clu
d
in
g
in
th
e
d
etec
tio
n
an
d
class
if
icatio
n
o
f
b
r
ain
tu
m
o
r
s
.
Ho
wev
er
,
th
e
ac
cu
r
ac
y
o
f
C
NNs
in
b
r
ain
tu
m
o
r
class
if
icatio
n
ca
n
s
till
b
e
im
p
r
o
v
ed
b
y
ap
p
l
y
in
g
ad
v
a
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
an
d
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
[
2
2
]
,
[
2
3
]
.
B
ased
o
n
th
e
liter
atu
r
e
r
ev
iew
o
u
tlin
e
d
ab
o
v
e,
th
is
s
tu
d
y
o
f
f
er
s
a
s
o
lu
tio
n
b
y
o
p
tim
izin
g
th
e
b
r
ain
tu
m
o
r
class
if
icatio
n
p
r
o
ce
s
s
th
r
o
u
g
h
th
e
ap
p
licat
io
n
o
f
ad
v
an
ce
d
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
an
d
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
[
2
4
]
,
[
2
5
]
.
Pre
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
s
u
ch
as
n
o
r
m
aliza
tio
n
,
d
ata
au
g
m
en
tatio
n
,
an
d
n
o
is
e
r
em
o
v
al
aim
to
e
n
h
an
ce
th
e
q
u
ality
o
f
MRI
im
ag
es
b
ef
o
r
e
th
ey
ar
e
p
r
o
ce
s
s
ed
b
y
th
e
m
o
d
el.
T
h
is
p
r
o
ce
s
s
r
esu
lts
in
m
o
r
e
r
ep
r
esen
tativ
e
f
ea
tu
r
es,
h
elp
i
n
g
th
e
m
o
d
el
b
etter
u
n
d
er
s
tan
d
r
elev
a
n
t
p
atter
n
s
[
1
9
]
,
[
2
6
]
.
On
t
h
e
o
t
h
er
h
a
n
d
,
e
n
s
em
b
le
lear
n
in
g
m
eth
o
d
s
s
u
ch
as
B
ag
g
in
g
ar
e
u
s
ed
to
im
p
r
o
v
e
t
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
class
if
icatio
n
m
o
d
el
b
y
c
o
m
b
in
i
n
g
p
r
ed
ictio
n
s
f
r
o
m
m
u
ltip
le
b
ase
m
o
d
els,
th
u
s
r
e
d
u
cin
g
th
e
r
is
k
o
f
o
v
e
r
f
itti
n
g
an
d
im
p
r
o
v
in
g
o
v
er
all
ac
cu
r
ac
y
.
B
y
i
n
teg
r
atin
g
d
ee
p
lear
n
in
g
ar
ch
i
tectu
r
es
lik
e
L
eNe
t
with
en
s
em
b
le
tech
n
i
q
u
es,
th
is
s
tu
d
y
aim
s
to
cr
ea
te
a
m
o
r
e
ac
cu
r
ate,
r
eliab
le,
a
n
d
e
f
f
i
cien
t
class
if
icatio
n
s
y
s
tem
[2
7
]
,
[2
8
]
.
T
h
is
s
tu
d
y
aim
s
to
d
ev
elo
p
a
n
MRI
-
b
ased
b
r
ain
tu
m
o
r
class
if
icatio
n
m
o
d
el
with
h
ig
h
a
cc
u
r
ac
y
b
y
o
p
tim
izin
g
th
e
co
m
b
in
atio
n
o
f
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
an
d
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
.
Sp
ec
if
ically
,
th
is
r
esear
ch
s
ee
k
s
to
an
aly
ze
th
e
im
p
ac
t
o
f
p
r
e
p
r
o
ce
s
s
in
g
t
ec
h
n
iq
u
es
o
n
m
o
d
el
p
er
f
o
r
m
an
ce
,
ev
alu
ate
th
e
ef
f
ec
tiv
en
ess
o
f
B
ag
g
in
g
m
et
h
o
d
s
in
im
p
r
o
v
in
g
class
if
icati
o
n
ac
cu
r
ac
y
,
an
d
id
en
tify
t
h
e
b
est
co
n
f
ig
u
r
atio
n
b
etwe
en
d
ee
p
lear
n
in
g
m
o
d
els
an
d
e
n
s
em
b
le
lear
n
in
g
m
eth
o
d
s
.
T
h
e
ex
p
ec
ted
f
i
n
al
o
u
tco
m
e
is
th
e
cr
ea
tio
n
o
f
a
s
y
s
tem
th
at
n
o
t
o
n
ly
ex
ce
ls
in
ac
cu
r
ac
y
b
u
t
also
ca
n
b
e
p
r
ac
tically
im
p
lem
en
ted
t
o
s
u
p
p
o
r
t
m
ed
ical
d
iag
n
o
s
es,
co
n
tr
ib
u
tin
g
t
o
t
h
e
ea
r
ly
d
etec
tio
n
o
f
b
r
ain
tu
m
o
r
s
an
d
im
p
r
o
v
in
g
p
atie
n
t
ca
r
e.
Ad
v
a
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
s
u
c
h
as
im
ag
e
n
o
r
m
aliza
tio
n
,
d
ata
au
g
m
en
tatio
n
,
an
d
n
o
is
e
r
em
o
v
al
ca
n
h
elp
im
p
r
o
v
e
th
e
q
u
ality
o
f
in
p
u
t
to
th
e
m
o
d
el,
r
esu
ltin
g
in
m
o
r
e
r
ep
r
esen
tativ
e
f
ea
tu
r
es.
On
th
e
o
th
er
h
a
n
d
,
en
s
em
b
le
lear
n
in
g
m
et
h
o
d
s
li
k
e
B
ag
g
in
g
o
f
f
er
an
a
p
p
r
o
ac
h
to
co
m
b
i
n
e
p
r
ed
ictio
n
s
f
r
o
m
s
ev
er
al
m
o
d
els
to
im
p
r
o
v
e
ac
cu
r
ac
y
an
d
r
e
d
u
ce
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
.
Ho
w
ev
er
,
th
e
ap
p
licatio
n
o
f
th
ese
m
eth
o
d
s
i
n
b
r
ai
n
tu
m
o
r
class
if
icatio
n
h
as
n
o
t
b
ee
n
f
u
lly
o
p
tim
ized
,
p
ar
ticu
la
r
ly
in
co
m
b
in
in
g
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
with
en
s
em
b
le
lear
n
in
g
a
p
p
r
o
ac
h
es
[
6
]
.
T
h
e
r
ef
o
r
e,
th
is
s
tu
d
y
ai
m
s
to
d
e
v
elo
p
a
m
o
r
e
ac
c
u
r
at
e
MRI
-
b
ased
b
r
ain
tu
m
o
r
class
if
icatio
n
m
o
d
el
b
y
o
p
tim
izin
g
p
r
e
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
an
d
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
.
T
h
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.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
0
6
-
5
1
1
9
5108
r
esu
ltin
g
m
o
d
el
is
ex
p
ec
ted
to
n
o
t
o
n
ly
im
p
r
o
v
e
class
if
icati
o
n
ac
cu
r
ac
y
b
u
t
also
s
u
p
p
o
r
t
m
ed
ical
p
r
ac
tice
in
p
r
o
v
id
i
n
g
f
aster
a
n
d
m
o
r
e
r
eli
ab
le
d
iag
n
o
s
es.
2.
M
E
T
H
O
D
T
h
is
s
tu
d
y
em
p
lo
y
s
a
q
u
an
tit
ativ
e
ex
p
er
im
en
tal
ap
p
r
o
ac
h
t
o
d
ev
elo
p
a
n
d
co
m
p
ar
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
b
r
ain
tu
m
o
r
class
if
icatio
n
m
o
d
els
u
s
in
g
v
ar
io
u
s
d
ee
p
l
ea
r
n
in
g
ar
ch
itectu
r
es.
T
h
e
p
r
im
ar
y
f
o
cu
s
o
f
th
e
r
esear
ch
is
to
c
o
m
p
ar
e
th
e
p
r
o
p
o
s
ed
R
esNet+Bag
g
in
g
m
o
d
el
with
in
d
i
v
id
u
al
m
o
d
els
s
u
ch
as
Go
o
g
L
eNe
t,
R
esNet,
L
eNe
t,
an
d
VGG
-
1
6
.
E
x
p
er
im
en
ts
ar
e
co
n
d
u
cted
u
s
in
g
a
b
r
ain
tu
m
o
r
MRI
im
ag
e
d
ataset
th
at
h
as
u
n
d
er
g
o
n
e
p
r
ep
r
o
ce
s
s
in
g
to
en
s
u
r
e
th
e
q
u
ality
o
f
th
e
i
n
p
u
t
d
a
ta
.
2
.
1
.
Resea
rc
h
d
a
t
a
s
et
T
h
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
co
n
s
is
ts
o
f
b
r
ain
tu
m
o
r
MRI
im
ag
es,
wh
ich
in
clu
d
e
m
ed
ical
im
ag
es
o
f
v
ar
io
u
s
ty
p
es
o
f
b
r
ain
tu
m
o
r
s
.
T
h
e
d
ata
is
s
o
u
r
ce
d
f
r
o
m
Kag
g
le
o
n
th
e
web
s
ite:
h
tt
p
s
:
//www
.
k
a
g
g
le
.
c
o
m
/c
o
d
e
/
g
u
s
lo
v
esm
at
h
/c
n
n
-
b
r
a
in
-
t
u
m
o
r
-
cl
ass
i
f
ic
ati
o
n
-
99
-
ac
c
u
r
ac
y
/i
n
p
u
t
wit
h
3
v
ar
ia
b
les
as
th
e
t
r
ai
n
i
n
g
m
o
d
e
l,
w
it
h
a
to
t
al
o
f
1
,
2
0
0
d
at
a.
T
h
is
d
atas
et
t
y
p
ica
ll
y
c
o
m
p
r
is
es
a
s
e
r
i
es
o
f
h
i
g
h
-
r
eso
lu
ti
o
n
im
a
g
es
,
e
n
a
b
l
in
g
i
n
-
d
e
p
t
h
p
r
o
ce
s
s
i
n
g
a
n
d
a
n
al
y
s
is
t
o
id
e
n
ti
f
y
p
att
er
n
s
i
n
t
h
e
b
r
ai
n
tu
m
o
r
s
.
T
h
es
e
MRI
im
ag
es
ar
e
g
e
n
er
ate
d
u
s
i
n
g
m
a
g
n
et
ic
r
es
o
n
an
ce
i
m
a
g
i
n
g
,
w
h
i
c
h
cl
ea
r
l
y
d
is
p
l
ay
s
t
h
e
d
e
tai
le
d
s
tr
u
ct
u
r
e
o
f
t
h
e
b
r
ai
n
.
T
y
p
i
ca
l
ly
,
t
h
e
d
a
tase
t
is
la
b
ele
d
as
h
e
m
o
r
r
h
a
g
e
a
n
d
n
o
n
-
h
em
o
r
r
h
ag
e,
w
h
ic
h
f
ac
i
lit
ate
s
t
h
e
t
r
a
in
in
g
a
n
d
ev
al
u
at
io
n
o
f
cl
ass
i
f
ic
ati
o
n
m
o
d
els
,
as
s
h
o
w
n
i
n
F
ig
u
r
e
1
.
Fig
u
r
e
1
s
h
o
ws
im
ag
es
in
t
h
e
d
ataset
th
at
v
ar
y
in
ter
m
s
o
f
tu
m
o
r
s
ize,
in
ten
s
ity
,
a
n
d
s
h
ap
e,
m
ak
in
g
p
r
ep
r
o
ce
s
s
in
g
ess
en
ti
al
to
en
h
an
ce
th
e
q
u
ality
o
f
in
p
u
t
f
o
r
d
ee
p
lea
r
n
in
g
m
o
d
els.
Sev
er
al
p
u
b
lic
d
atasets
u
s
ed
in
th
is
s
tu
d
y
in
clu
d
e
th
e
b
r
ain
MRI
im
ag
es
d
ataset
av
ailab
le
o
n
Kag
g
le
a
n
d
th
e
b
r
ai
n
tu
m
o
r
s
eg
m
en
tatio
n
ch
allen
g
e
(
B
R
AT
S)
(
d
ataset,
wh
ich
ar
e
f
r
eq
u
en
tly
u
tili
ze
d
in
r
esear
ch
r
elate
d
to
b
r
ain
tu
m
o
r
s
eg
m
en
tatio
n
an
d
class
if
icatio
n
.
T
h
ese
d
atasets
co
n
tain
a
w
id
e
r
an
g
e
o
f
im
a
g
es,
co
v
er
in
g
v
ar
io
u
s
s
tag
es
o
f
tu
m
o
r
d
ev
el
o
p
m
en
t
an
d
p
atien
t
co
n
d
itio
n
s
,
en
ab
lin
g
th
e
m
o
d
el
to
lear
n
to
r
ec
o
g
n
ize
th
e
d
iv
er
s
e
p
atter
n
s
p
r
esen
t
in
b
r
ain
tu
m
o
r
s
.
T
h
e
u
s
e
o
f
h
ig
h
-
q
u
ality
d
atasets
with
r
ich
v
ar
iatio
n
is
cr
u
cial
f
o
r
im
p
r
o
v
i
n
g
th
e
m
o
d
el'
s
ab
ilit
y
to
ac
cu
r
ately
cl
ass
if
y
b
r
ain
tu
m
o
r
MRI
im
ag
e
s
.
T
h
e
MRI
im
a
g
es
s
h
o
w
two
ca
teg
o
r
ies
o
f
b
r
ai
n
co
n
d
itio
n
s
,
n
am
el
y
h
em
o
r
r
h
a
g
e
as
s
h
o
wn
in
Fig
u
r
e
1
(
a)
an
d
n
o
n
-
h
em
o
r
r
h
ag
e
as
s
h
o
wn
in
Fig
u
r
e
1
(
b
)
.
I
n
ca
teg
o
r
y
(
a
)
h
em
o
r
r
h
ag
e
(
to
p
r
o
w)
,
th
er
e
is
an
ar
ea
with
b
r
ig
h
t
in
ten
s
ity
(
wh
ite)
in
d
icatin
g
b
lo
o
d
ac
cu
m
u
latio
n
d
u
e
to
b
lee
d
in
g
,
u
s
u
ally
ir
r
eg
u
lar
in
s
h
ap
e
ac
co
r
d
in
g
to
th
e
lo
ca
tio
n
an
d
ca
u
s
e.
Me
an
wh
ile,
ca
teg
o
r
y
(
b
)
n
o
n
-
h
e
m
o
r
r
h
ag
e
(
b
o
tto
m
r
o
w)
s
h
o
ws
an
MRI
o
f
th
e
b
r
ain
with
o
u
t
in
d
icatio
n
o
f
b
leed
i
n
g
,
with
m
o
r
e
u
n
i
f
o
r
m
p
ix
el
in
ten
s
ity
an
d
n
o
r
m
al
-
lo
o
k
in
g
b
r
ain
s
tr
u
ctu
r
es,
alth
o
u
g
h
it
m
a
y
s
till
in
d
icate
o
th
er
ab
n
o
r
m
alities
.
T
h
ese
two
ca
teg
o
r
ies
p
lay
an
im
p
o
r
tan
t
r
o
le
i
n
h
elp
in
g
clin
ical
d
iag
n
o
s
is
to
d
eter
m
in
e
th
e
ty
p
e
o
f
b
r
ai
n
d
is
o
r
d
e
r
an
d
ap
p
r
o
p
r
i
ate
tr
ea
tm
en
t.
(
a)
(
b
)
Fig
u
r
e
1
.
Sam
p
le
b
r
ain
t
u
m
o
r
MRI
(
a)
h
em
o
r
r
h
ag
e
,
an
d
(
b
)
n
o
n
-
h
em
o
r
r
h
ag
e
2
.
2
.
P
r
o
po
s
ed
m
o
del
I
n
th
e
p
r
o
p
o
s
ed
R
esNet+
B
ag
g
in
g
m
o
d
el,
B
ag
g
in
g
(
b
o
o
ts
tr
ap
ag
g
r
eg
atin
g
)
is
ap
p
lied
t
o
en
h
an
ce
m
o
d
el
p
er
f
o
r
m
an
ce
b
y
c
o
m
b
in
in
g
p
r
e
d
ictio
n
s
f
r
o
m
m
u
l
tip
le
in
d
ep
e
n
d
en
tly
tr
ain
e
d
R
esNet
m
o
d
els
o
n
d
if
f
er
en
t
s
u
b
s
ets
o
f
th
e
d
ata.
T
h
e
co
r
e
co
n
ce
p
t
o
f
B
ag
g
i
n
g
is
to
g
en
e
r
ate
s
ev
er
al
m
o
d
els
b
y
r
a
n
d
o
m
l
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
Op
timiz
in
g
b
r
a
in
tu
mo
r
MRI
cla
s
s
ifica
tio
n
u
s
in
g
a
d
va
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
… (
A
kim
Ma
n
a
o
r
Ha
r
a
P
a
r
d
ed
e
)
5109
s
am
p
lin
g
f
r
o
m
th
e
tr
ain
i
n
g
d
ataset
u
s
in
g
b
o
o
ts
tr
ap
s
am
p
lin
g
,
wh
ich
allo
ws
ea
ch
m
o
d
e
l
to
lear
n
f
r
o
m
a
d
if
f
er
en
t
s
u
b
s
et
o
f
d
ata.
E
ac
h
R
esNet
m
o
d
el
p
r
o
v
id
es
an
i
n
d
ep
en
d
en
t
p
r
ed
ictio
n
,
wh
ic
h
is
th
en
co
m
b
i
n
ed
u
s
in
g
m
ajo
r
ity
v
o
tin
g
f
o
r
cl
ass
if
icatio
n
.
Hy
p
er
p
ar
am
eter
s
:
lear
n
in
g
r
ate=
0
.
0
0
1
,
o
p
tim
izer
=A
d
am
,
b
atch
s
ize=
3
2
,
ep
o
ch
s
=5
0
.
Nu
m
b
e
r
o
f
R
esNet
b
ase
lear
n
er
s
i
n
B
ag
g
in
g
=5
.
T
h
is
ap
p
r
o
ac
h
aim
s
to
im
p
r
o
v
e
g
en
er
aliza
tio
n
an
d
r
ed
u
ce
t
h
e
lik
elih
o
o
d
o
f
o
v
er
f
itti
n
g
b
y
d
ec
r
ea
s
in
g
th
e
v
a
r
ian
ce
o
f
in
d
iv
id
u
al
m
o
d
el
p
r
ed
ictio
n
s
.
B
ag
g
in
g
ap
p
lied
to
R
esNet
h
elp
s
im
p
r
o
v
e
s
tab
ilit
y
an
d
ac
cu
r
ac
y
,
p
a
r
ticu
lar
ly
wh
en
d
ea
lin
g
with
s
u
b
tle
d
if
f
er
en
ce
s
b
etwe
en
m
o
d
els,
wh
ich
is
o
f
ten
en
co
u
n
ter
ed
in
m
ed
ical
im
ag
e
class
if
icatio
n
.
T
ab
le
1
p
r
esen
ts
a
co
m
p
ar
is
o
n
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
with
o
t
h
er
m
o
d
els.
I
n
T
ab
le
1
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el,
th
e
d
ata
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
p
lay
s
a
cr
u
cial
r
o
le
in
en
s
u
r
in
g
th
e
q
u
ality
o
f
th
e
b
r
ai
n
tu
m
o
r
M
R
I
im
ag
es
u
s
ed
f
o
r
m
o
d
el
tr
ain
in
g
.
T
h
is
p
r
ep
r
o
ce
s
s
in
g
p
r
o
c
ess
in
v
o
lv
es
s
ev
er
al
tech
n
iq
u
es,
s
u
ch
as
im
ag
e
n
o
r
m
aliza
tio
n
,
wh
ich
s
ca
les
th
e
p
ix
el
v
alu
es
to
th
e
r
an
g
e
[
0
,
1
]
to
im
p
r
o
v
e
m
o
d
el
co
n
v
er
g
en
ce
,
a
n
d
d
ata
a
u
g
m
en
tatio
n
,
in
clu
d
in
g
r
o
tatio
n
,
f
lip
p
in
g
,
zo
o
m
in
g
,
a
n
d
cr
o
p
p
in
g
,
aim
ed
at
ex
p
an
d
i
n
g
th
e
d
ataset’
s
v
ar
iat
io
n
an
d
en
h
an
cin
g
th
e
m
o
d
el’
s
ab
ilit
y
to
g
e
n
er
alize
.
A
d
d
itio
n
ally
,
th
e
im
ag
es
ar
e
r
esized
to
m
atch
th
e
m
o
d
e
l's
in
p
u
t
lay
er
,
s
u
c
h
as
2
2
4
×2
2
4
f
o
r
R
esNet
an
d
2
8
×2
8
f
o
r
L
eNe
t.
An
o
th
er
s
tep
tak
en
is
n
o
is
e
r
ed
u
ctio
n
,
to
r
em
o
v
e
ir
r
elev
a
n
t
d
is
tu
r
b
an
c
es
in
th
e
im
ag
es
th
at
co
u
ld
r
ed
u
ce
th
e
m
o
d
el
’
s
ac
cu
r
ac
y
.
All
o
f
t
h
ese
s
tep
s
ar
e
d
esig
n
ed
to
e
n
s
u
r
e
th
at
t
h
e
d
ata
u
s
ed
is
o
f
h
i
g
h
q
u
ality
an
d
r
ea
d
y
to
b
e
p
r
o
ce
s
s
ed
b
y
t
h
e
m
o
d
el,
th
u
s
i
m
p
r
o
v
i
n
g
its
ab
ilit
y
to
ac
cu
r
at
ely
class
if
y
b
r
ain
tu
m
o
r
s
.
T
h
e
B
ag
g
in
g
tech
n
iq
u
e
is
ap
p
lied
b
y
co
n
s
tr
u
ctin
g
m
u
ltip
le
R
esNet
m
o
d
els th
at
ar
e
tr
ain
ed
i
n
p
ar
allel
o
n
d
if
f
e
r
en
t
d
ata
s
u
b
s
ets.
E
ac
h
m
o
d
el
p
r
o
v
id
es
its
p
r
ed
ic
tio
n
,
wh
ich
is
th
en
co
m
b
i
n
e
d
th
r
o
u
g
h
m
ajo
r
ity
v
o
tin
g
t
o
d
ete
r
m
in
e
t
h
e
f
in
al
o
u
tco
m
e.
T
h
is
ap
p
r
o
ac
h
aim
s
to
im
p
r
o
v
e
g
e
n
er
aliza
tio
n
a
n
d
r
ed
u
ce
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
.
B
ag
g
in
g
en
h
a
n
ce
s
R
esNet
'
s
ab
ilit
y
to
r
ec
o
g
n
ize
f
ea
tu
r
es
an
d
p
atter
n
s
in
b
r
ai
n
tu
m
o
r
MRI
im
a
g
es
b
y
r
e
d
u
cin
g
r
elian
ce
o
n
a
s
in
g
le
m
o
d
el
th
at
m
a
y
b
e
p
r
o
n
e
to
o
v
er
f
itti
n
g
.
B
y
co
m
b
in
in
g
s
ev
er
al
R
esNet
m
o
d
els,
th
e
f
in
al
class
if
icatio
n
r
esu
lt
b
ec
o
m
es
m
o
r
e
ac
cu
r
at
e
as
it
is
les
s
in
f
lu
en
ce
d
b
y
er
r
o
r
s
o
r
n
o
is
e
with
in
s
p
ec
if
ic
d
ata
s
u
b
s
ets.
I
n
co
m
p
ar
is
o
n
,
t
h
e
R
esNet
m
o
d
el
with
o
u
t
B
ag
g
i
n
g
is
m
o
r
e
li
k
ely
to
s
u
f
f
er
f
r
o
m
o
v
er
f
itti
n
g
if
th
e
d
ataset
is
n
o
t
lar
g
e
o
r
d
iv
er
s
e
e
n
o
u
g
h
,
d
esp
ite
th
e
ar
c
h
itectu
r
e'
s
s
tr
en
g
th
in
ca
p
tu
r
in
g
f
ea
tu
r
es
f
r
o
m
im
a
g
es
d
u
e
to
it
s
g
r
ea
ter
d
ep
th
.
T
h
e
R
esNet+
B
ag
g
in
g
m
o
d
el
is
ex
p
ec
te
d
to
p
er
f
o
r
m
b
etter
th
an
o
th
er
m
o
d
els,
s
u
ch
as
R
esNet,
Go
o
g
L
eNe
t,
VGG
-
1
6
,
an
d
L
eNe
t,
d
u
e
to
th
e
co
m
b
in
ed
s
tr
en
g
th
o
f
R
esNet
in
ex
tr
ac
tin
g
co
m
p
le
x
f
ea
tu
r
es
f
r
o
m
im
ag
es
an
d
th
e
s
tab
ilit
y
p
r
o
v
id
ed
b
y
B
ag
g
in
g
.
B
y
r
ed
u
cin
g
p
r
ed
ictio
n
v
ar
ia
n
ce
a
n
d
ad
d
r
e
s
s
in
g
o
v
er
f
itti
n
g
,
th
is
m
o
d
el
i
s
m
o
r
e
ef
f
ec
tiv
e
in
h
an
d
lin
g
d
iv
er
s
e
an
d
co
m
p
lex
m
ed
ical
d
atasets
.
T
ab
le
1
.
Mo
d
el
co
m
p
ar
is
on
M
o
d
e
l
M
e
t
h
o
d
M
a
i
n
o
b
j
e
c
t
i
v
e
P
o
t
e
n
t
i
a
l
u
se
R
e
sN
e
t
+
B
a
g
g
i
n
g
R
e
sN
e
t
+
B
a
g
g
i
n
g
(
e
n
s
e
m
b
l
e
)
I
mp
r
o
v
e
a
c
c
u
r
a
c
y
a
n
d
g
e
n
e
r
a
l
i
z
a
t
i
o
n
b
y
c
o
m
b
i
n
i
n
g
m
u
l
t
i
p
l
e
R
e
sN
e
t
mo
d
e
l
s
,
r
e
d
u
c
i
n
g
o
v
e
r
f
i
t
t
i
n
g
,
a
n
d
e
n
h
a
n
c
i
n
g
s
t
a
b
i
l
i
t
y
.
Ef
f
e
c
t
i
v
e
f
o
r
l
a
r
g
e
a
n
d
c
o
m
p
l
e
x
d
a
t
a
se
t
s
s
u
c
h
a
s
b
r
a
i
n
t
u
mo
r
M
R
I
i
ma
g
e
s,
o
f
f
e
r
i
n
g
h
i
g
h
st
a
b
i
l
i
t
y
a
n
d
a
c
c
u
r
a
c
y
.
R
e
sN
e
t
R
e
si
d
u
a
l
n
e
t
w
o
r
k
s
(
D
e
e
p
C
N
N
)
A
d
d
r
e
ss
t
h
e
v
a
n
i
sh
i
n
g
g
r
a
d
i
e
n
t
p
r
o
b
l
e
m
a
n
d
r
e
c
o
g
n
i
z
e
c
o
m
p
l
e
x
f
e
a
t
u
r
e
s
i
n
i
m
a
g
e
s.
S
u
i
t
a
b
l
e
f
o
r
l
a
r
g
e
d
a
t
a
set
s
a
n
d
med
i
c
a
l
i
m
a
g
e
s
r
e
q
u
i
r
i
n
g
i
n
-
d
e
p
t
h
a
n
a
l
y
s
i
s,
b
u
t
p
r
o
n
e
t
o
o
v
e
r
f
i
t
t
i
n
g
w
i
t
h
sma
l
l
d
a
t
a
set
s.
G
o
o
g
L
e
N
e
t
I
n
c
e
p
t
i
o
n
(
mu
l
t
i
-
sca
l
e
c
o
n
v
o
l
u
t
i
o
n
a
l
f
i
l
t
e
r
s
)
R
e
d
u
c
e
t
h
e
n
u
m
b
e
r
o
f
p
a
r
a
met
e
r
s
a
n
d
c
a
p
t
u
r
e
m
u
l
t
i
-
s
c
a
l
e
i
n
f
o
r
ma
t
i
o
n
.
I
d
e
a
l
f
o
r
a
p
p
l
i
c
a
t
i
o
n
s
w
i
t
h
c
o
m
p
u
t
a
t
i
o
n
a
l
l
i
m
i
t
a
t
i
o
n
s,
s
u
c
h
a
s
me
d
i
c
a
l
i
m
a
g
e
s,
r
e
q
u
i
r
i
n
g
e
f
f
i
c
i
e
n
c
y
w
i
t
h
o
u
t
sac
r
i
f
i
c
i
n
g
a
c
c
u
r
a
c
y
.
VGG
-
16
D
e
e
p
C
N
N
(
m
u
l
t
i
p
l
e
c
o
n
v
o
l
u
t
i
o
n
a
l
l
a
y
e
r
s)
R
e
c
o
g
n
i
z
e
v
i
su
a
l
p
a
t
t
e
r
n
s
t
h
r
o
u
g
h
c
o
n
se
c
u
t
i
v
e
c
o
n
v
o
l
u
t
i
o
n
a
l
l
a
y
e
r
s.
S
u
i
t
a
b
l
e
f
o
r
sma
l
l
e
r
o
r
me
d
i
u
m
-
si
z
e
d
d
a
t
a
s
e
t
s
b
u
t
c
a
n
su
f
f
e
r
f
r
o
m
o
v
e
r
f
i
t
t
i
n
g
w
i
t
h
o
u
t
p
r
o
p
e
r
o
p
t
i
m
i
z
a
t
i
o
n
t
e
c
h
n
i
q
u
e
s
.
Le
N
e
t
S
i
mp
l
e
C
N
N
(
C
o
n
v
o
l
u
t
i
o
n
a
l
a
n
d
s
u
b
sam
p
l
i
n
g
l
a
y
e
r
s
)
A
l
i
g
h
t
w
e
i
g
h
t
a
n
d
si
m
p
l
e
m
o
d
e
l
su
i
t
a
b
l
e
f
o
r
sm
a
l
l
d
a
t
a
s
e
t
s
a
n
d
a
p
p
l
i
c
a
t
i
o
n
s
t
h
a
t
d
o
n
’
t
r
e
q
u
i
r
e
h
i
g
h
c
o
m
p
l
e
x
i
t
y
.
I
d
e
a
l
f
o
r
sm
a
l
l
d
a
t
a
se
t
s
a
n
d
a
p
p
l
i
c
a
t
i
o
n
s
w
i
t
h
l
i
m
i
t
e
d
r
e
so
u
r
c
e
s
,
b
u
t
l
e
ss
e
f
f
e
c
t
i
v
e
f
o
r
c
o
m
p
l
e
x
me
d
i
c
a
l
i
ma
g
e
s
.
2
.
3
.
Resea
rc
h
f
ra
m
ewo
r
k
T
h
e
r
esear
ch
f
r
am
ewo
r
k
i
n
th
i
s
s
tu
d
y
is
cr
u
cial
as it p
r
o
v
id
e
s
a
s
y
s
tem
atic
g
u
id
e
f
o
r
ea
ch
s
tag
e
o
f
th
e
r
esear
ch
,
f
r
o
m
d
ata
p
r
o
ce
s
s
in
g
to
r
esu
lt
e
v
alu
atio
n
.
I
n
th
is
s
tu
d
y
,
th
e
f
r
am
ewo
r
k
h
elp
s
o
r
g
an
ize
th
e
p
r
o
ce
s
s
o
f
co
m
p
ar
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
v
ar
io
u
s
C
NN
m
o
d
el
ar
ch
itectu
r
es,
s
u
ch
as
L
eNe
t,
R
e
s
Net,
Go
o
g
L
eNe
t,
VGGN
et,
an
d
th
e
p
r
o
p
o
s
ed
m
o
d
el
(
R
esNet+Bag
g
in
g
)
.
W
ith
a
clea
r
f
r
am
ewo
r
k
,
th
e
r
esear
ch
er
ca
n
p
er
f
o
r
m
co
m
p
ar
is
o
n
s
tr
an
s
p
ar
e
n
tly
a
n
d
ac
cu
r
ately
,
en
ab
lin
g
a
b
etter
u
n
d
e
r
s
tan
d
in
g
o
f
ea
c
h
r
esear
ch
p
h
ase.
Fu
r
th
er
m
o
r
e
,
th
is
f
r
am
ew
o
r
k
m
in
im
izes
er
r
o
r
s
a
n
d
e
n
s
u
r
e
s
a
co
m
p
r
e
h
en
s
iv
e
ev
alu
atio
n
b
y
u
s
in
g
v
ar
io
u
s
p
er
f
o
r
m
an
ce
m
etr
ics,
m
a
k
in
g
t
h
e
r
esear
ch
r
esu
lts
m
o
r
e
v
alid
an
d
r
eliab
le.
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
0
6
-
5
1
1
9
5110
T
h
e
r
e
s
e
a
r
c
h
f
r
a
m
ew
o
r
k
i
ll
u
s
t
r
a
t
e
d
i
n
F
i
g
u
r
e
2
o
u
t
li
n
e
s
t
h
e
w
o
r
k
f
l
o
w
,
b
e
g
i
n
n
i
n
g
wi
t
h
b
r
a
i
n
M
R
I
i
m
a
g
e
s
d
a
t
as
e
t
,
w
h
ic
h
c
o
n
s
i
s
t
s
o
f
t
w
o
c
l
a
s
s
es
:
h
e
m
o
r
r
h
a
g
e
a
n
d
n
o
n
-
h
e
m
o
r
r
h
a
g
e
.
T
h
e
p
r
o
c
e
s
s
s
t
a
r
ts
w
it
h
d
a
t
a
p
r
e
p
r
o
c
e
s
s
i
n
g
,
w
h
e
r
e
i
m
a
g
e
d
a
t
a
i
s
e
n
h
a
n
c
e
d
t
h
r
o
u
g
h
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
t
e
c
h
n
i
q
u
e
s
s
u
c
h
as
f
l
i
p
p
i
n
g
,
r
o
t
a
t
i
n
g
,
s
h
e
a
r
i
n
g
,
a
n
d
r
e
s
c
a
li
n
g
.
T
h
e
s
e
t
e
c
h
n
i
q
u
e
s
a
i
m
t
o
i
n
c
r
e
as
e
t
h
e
v
a
r
i
e
t
y
o
f
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
,
u
l
ti
m
a
t
e
l
y
i
m
p
r
o
v
i
n
g
t
h
e
m
o
d
e
l'
s
a
b
il
i
t
y
t
o
r
e
c
o
g
n
i
ze
a
b
r
o
a
d
e
r
r
a
n
g
e
o
f
p
a
t
t
e
r
n
s
.
F
o
l
l
o
w
i
n
g
p
r
e
p
r
o
c
e
s
s
i
n
g
,
t
h
e
d
a
t
a
s
p
l
i
tt
i
n
g
s
t
a
g
e
d
i
v
i
d
e
s
t
h
e
d
at
a
s
e
t
i
n
t
o
t
h
r
e
e
s
u
b
s
e
t
s
:
8
0
%
f
o
r
t
r
ai
n
i
n
g
,
1
0
%
f
o
r
v
a
l
i
d
a
t
i
o
n
t
o
as
s
e
s
s
t
h
e
m
o
d
e
l
’
s
p
e
r
f
o
r
m
a
n
c
e
d
u
r
i
n
g
t
r
a
i
n
i
n
g
,
a
n
d
1
0
%
f
o
r
t
es
t
i
n
g
,
u
s
e
d
t
o
e
v
al
u
a
t
e
t
h
e
m
o
d
e
l'
s
f
i
n
a
l
p
e
r
f
o
r
m
a
n
c
e
a
f
t
e
r
t
r
ain
i
n
g
.
A
t
t
h
is
s
t
a
g
e,
d
a
t
a
v
i
s
u
a
li
z
a
ti
o
n
i
s
a
ls
o
p
e
r
f
o
r
m
e
d
t
o
v
e
r
i
f
y
t
h
e
d
i
s
t
r
i
b
u
t
i
o
n
o
f
d
a
t
a
a
c
r
o
s
s
t
h
e
d
i
f
f
e
r
e
n
t
c
la
s
s
e
s
.
I
n
t
h
e
m
o
d
e
l
t
r
a
i
n
i
n
g
a
n
d
cl
a
s
s
i
f
i
c
a
t
i
o
n
p
h
a
s
e
,
t
h
e
p
r
e
p
r
o
c
e
s
s
e
d
d
at
a
is
u
s
e
d
t
o
t
r
ai
n
v
a
r
i
o
u
s
m
o
d
e
l
a
r
c
h
i
t
e
c
t
u
r
es
,
i
n
cl
u
d
i
n
g
L
e
N
e
t
,
R
es
N
e
t
,
G
o
o
g
L
e
N
e
t
,
VG
G
N
et
,
a
s
w
e
l
l
a
s
t
h
e
p
r
o
p
o
s
ed
m
o
d
i
f
i
e
d
m
o
d
e
l
(
R
es
N
e
t+
B
ag
g
i
n
g
)
.
Fig
u
r
e
2
s
h
o
ws
th
e
r
esear
ch
d
esig
n
b
ased
o
n
all
m
o
d
els
tr
ain
ed
o
n
t
h
e
s
am
e
d
ataset
to
cr
ea
te
a
b
r
ain
tu
m
o
r
class
if
icatio
n
m
o
d
el.
Af
ter
tr
ain
in
g
,
th
e
m
o
d
els
u
n
d
e
r
g
o
a
m
o
d
el
co
m
p
ar
is
o
n
,
wh
e
r
e
th
e
p
er
f
o
r
m
an
ce
o
f
L
eNe
t,
R
esNet,
Go
o
g
L
eNe
t,
an
d
VGGN
et
is
co
m
p
ar
ed
with
th
e
p
r
o
p
o
s
ed
m
o
d
el.
T
h
i
s
co
m
p
ar
is
o
n
s
er
v
es
to
ev
alu
ate
th
e
im
p
ac
t
o
f
ar
c
h
itectu
r
al
m
o
d
if
icatio
n
s
an
d
h
y
p
er
p
a
r
am
eter
tu
n
in
g
in
im
p
r
o
v
in
g
class
if
icatio
n
ac
cu
r
ac
y
.
Fin
ally
,
in
t
h
e
ev
al
u
atio
n
an
d
a
n
aly
s
is
s
tag
e,
th
e
class
if
icatio
n
r
esu
lts
o
f
ea
ch
m
o
d
el
ar
e
ass
ess
ed
u
s
in
g
p
er
f
o
r
m
an
ce
m
etr
ics
s
u
ch
as
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
lo
s
s
,
s
p
ec
if
icity
,
Ma
tth
ews
co
r
r
elatio
n
c
o
ef
f
icien
t
(
MCC
)
,
an
d
ar
ea
u
n
d
e
r
th
e
r
ec
eiv
er
o
p
e
r
atin
g
c
h
ar
ac
ter
is
tic
c
u
r
v
e
(
AUC
-
R
O
C
)
.
T
h
e
r
esu
lts
ar
e
th
en
p
r
esen
ted
v
is
u
ally
to
f
ac
ilit
ate
in
ter
p
r
eta
tio
n
an
d
f
u
r
th
e
r
an
aly
s
is
.
Fig
u
r
e
2
.
R
esear
ch
f
r
am
ewo
r
k
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Da
t
a
prepro
ce
s
s
ing
a
nd
da
t
a
a
ug
m
ent
a
t
io
n pro
ce
s
s
I
n
th
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e,
t
h
e
b
r
ain
tu
m
o
r
MRI
im
ag
es
u
n
d
e
r
g
o
a
s
er
ies
o
f
d
ata
en
h
an
ce
m
en
t
p
r
o
ce
s
s
es
.
T
h
ese
p
r
o
ce
s
s
es
in
clu
d
e
:
f
lip
p
in
g
–
f
lip
p
in
g
th
e
i
m
ag
e
h
o
r
izo
n
tally
o
r
v
er
tically
to
cr
ea
te
v
ar
iatio
n
s
in
o
r
ien
tatio
n
;
r
o
tatio
n
–
r
o
tatin
g
th
e
im
ag
e
at
s
p
ec
if
ic
an
g
les
to
in
cr
ea
s
e
th
e
d
iv
er
s
ity
o
f
th
e
d
ataset;
s
h
ea
r
–
ap
p
ly
i
n
g
a
s
h
ea
r
tr
an
s
f
o
r
m
atio
n
to
cr
ea
te
s
lig
h
t
d
is
to
r
tio
n
s
in
th
e
im
ag
e;
an
d
r
escalin
g
–
n
o
r
m
alizin
g
th
e
p
ix
el
v
alu
es
to
a
[
0
,
1
]
r
an
g
e
to
h
elp
t
h
e
m
o
d
el
b
etter
r
e
co
g
n
ize
p
atter
n
s
in
th
e
d
ata.
T
h
e
r
esu
lts
o
f
th
e
au
g
m
en
tatio
n
p
r
o
ce
s
s
ca
n
b
e
s
ee
n
in
Fig
u
r
e
3
.
Fig
u
r
e
3
s
h
o
ws
th
e
r
esu
lts
o
f
th
e
d
ata
au
g
m
en
tatio
n
p
r
o
ce
s
s
ap
p
lied
to
th
e
b
r
ain
tu
m
o
r
MRI
s
ca
n
im
ag
es.
T
h
e
au
g
m
en
tatio
n
te
ch
n
iq
u
es,
s
u
ch
as
f
lip
p
in
g
,
r
o
tatin
g
,
an
d
r
escalin
g
,
s
ig
n
if
ican
tly
en
r
ich
t
h
e
d
ataset
with
v
ar
ied
s
am
p
les.
T
h
ese
tech
n
iq
u
es
ef
f
ec
tiv
e
ly
en
h
an
ce
th
e
m
o
d
el'
s
g
en
er
aliza
tio
n
ab
ilit
y
.
T
h
e
p
r
im
ar
y
g
o
al
o
f
au
g
m
e
n
t
atio
n
is
to
en
s
u
r
e
th
at
th
e
m
o
d
el
d
o
es
n
o
t
b
ec
o
m
e
o
v
e
r
ly
d
ep
en
d
en
t
o
n
s
p
ec
if
ic
p
atter
n
s
,
m
ak
in
g
it
m
o
r
e
ad
a
p
tab
le
to
d
iv
er
s
e
test
d
ata.
I
n
t
h
is
s
tu
d
y
,
th
e
ex
p
e
r
im
en
tal
r
e
s
u
lts
wer
e
o
b
tain
ed
f
r
o
m
class
if
y
in
g
MRI
b
r
ain
tu
m
o
r
im
ag
es,
wh
ich
wer
e
d
iv
id
ed
in
to
th
r
ee
ca
teg
o
r
i
es:
h
em
o
r
r
h
ag
e
an
d
non
-
h
em
o
r
r
h
ag
e.
T
h
e
d
ataset
u
n
d
e
r
wen
t
p
r
e
p
r
o
ce
s
s
in
g
,
i
n
clu
d
in
g
d
ata
au
g
m
en
tatio
n
w
ith
tech
n
iq
u
es
lik
e
f
lip
p
in
g
,
r
o
tatin
g
,
s
h
ea
r
i
n
g
,
a
n
d
r
escalin
g
.
Af
ter
p
r
ep
r
o
ce
s
s
in
g
,
th
e
d
ataset
was
s
p
lit
in
to
tr
ain
in
g
d
ata
(
8
0
%),
v
alid
atio
n
d
ata
(
1
0
%),
a
n
d
test
in
g
d
ata
(
1
0
%).
T
h
is
p
r
o
ce
s
s
en
s
u
r
es th
at
th
e
m
o
d
el
e
v
alu
ati
o
n
is
p
er
f
o
r
m
ed
o
n
u
n
s
ee
n
d
ata,
p
r
o
v
id
in
g
o
b
jecti
v
e
an
d
r
ea
lis
tic
r
esu
lts
.
P
r
e
pr
oc
e
s
s
i
ng
D
a
t
a
A
ugm
e
nt
a
t
i
on
F
l
i
p
p
i
n
g
R
o
t
a
t
e
M
R
I
I
ma
ge
O
f B
r
ai
n
Tu
mor
S
pl
i
t
D
a
t
a
T
r
a
i
n
i
n
g
D
a
t
a
0
.
80
V
a
l
i
d
a
t
i
o
n
D
a
t
a
0
.
10
P
re
pa
ri
ng
D
a
t
a
M
o
d
e
l
T
r
a
n
i
n
g
C
l
a
s
s
i
f
i
c
a
t
i
o
n
I
m
a
g
e
C
l
a
s
s
i
f
i
c
a
t
i
o
n
D
a
t
a
s
e
t
P
r
op
os
e
d M
e
t
ho
d
(
R
e
s
N
e
t
+
B
a
ggi
ng
)
D
a
t
a
V
i
s
ua
l
i
z
a
t
i
on
T
r
a
i
n
i
n
g
D
a
t
a
s
e
t
Cl
a
s
s
i
fi
c
a
t
i
on
re
s
ul
t
R
e
s
u
l
t
s
T
e
s
t
i
n
g
D
a
t
a
0
.
10
E
va
l
ua
t
i
on a
nd
A
na
l
ys
i
s
of
Re
s
ul
t
s
L
e
N
e
t
R
e
s
N
e
t
S
he
a
r
r
a
nge
R
e
s
c
a
l
e
V
g
g
N
e
t
G
o
o
g
l
e
N
e
t
M
od
e
l
C
om
p
ar
i
s
on
H
e
m
m
org
ha
ge
N
on
H
e
m
m
or
gha
ge
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
Op
timiz
in
g
b
r
a
in
tu
mo
r
MRI
cla
s
s
ifica
tio
n
u
s
in
g
a
d
va
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
… (
A
kim
Ma
n
a
o
r
Ha
r
a
P
a
r
d
ed
e
)
5111
Fig
u
r
e
3
.
Au
g
m
en
tatio
n
o
f
b
r
a
in
tu
m
o
r
MRI
im
a
g
e
d
ataset
3
.
2
.
M
o
del
t
ra
ini
ng
a
nd
cla
s
s
if
ica
t
io
n
S
e
v
e
r
a
l
C
N
N
m
o
d
e
l
s
–
L
e
N
e
t
,
R
e
s
N
e
t
,
G
o
o
g
L
e
N
e
t
,
V
G
G
N
e
t
,
a
n
d
t
h
e
p
r
o
p
o
s
e
d
m
o
d
e
l
,
w
e
r
e
e
v
a
l
u
a
t
e
d
f
o
r
b
r
a
i
n
t
u
m
o
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
u
s
i
n
g
M
R
I
i
m
a
g
e
s
.
E
a
c
h
m
o
d
e
l
e
x
h
i
b
i
t
e
d
d
i
f
f
e
r
e
n
t
l
e
v
e
l
s
o
f
a
c
c
u
r
a
c
y
a
n
d
l
e
a
r
n
i
n
g
b
e
h
a
v
i
o
r
d
u
e
t
o
t
h
e
i
r
a
r
c
h
i
t
e
c
t
u
r
a
l
d
i
f
f
e
r
e
n
c
e
s
.
W
h
i
l
e
s
i
m
p
l
e
r
m
o
d
e
l
s
l
i
k
e
L
e
N
e
t
s
h
o
w
e
d
l
i
m
i
t
e
d
p
e
r
f
o
r
m
a
n
c
e
,
d
e
e
p
e
r
m
o
d
e
l
s
s
u
c
h
a
s
R
e
s
N
e
t
,
G
o
o
g
L
e
N
e
t
,
a
n
d
V
G
G
N
e
t
a
c
h
i
e
v
e
d
b
e
t
t
e
r
r
e
s
u
l
t
s
o
w
i
n
g
t
o
t
h
e
i
r
a
d
v
a
n
c
e
d
s
t
r
u
c
t
u
r
e
s
.
T
h
e
p
r
o
p
o
s
e
d
m
o
d
e
l
,
d
e
s
i
g
n
e
d
s
p
e
c
i
f
i
c
a
l
l
y
f
o
r
t
h
i
s
t
a
s
k
,
a
i
m
e
d
t
o
o
p
t
i
m
i
z
e
b
o
t
h
a
c
c
u
r
a
c
y
a
n
d
c
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
c
i
e
n
c
y
.
F
i
g
u
r
e
4
s
h
o
w
s
t
h
e
t
r
a
i
n
i
n
g
a
c
c
u
r
a
c
y
c
u
r
v
e
s
,
i
l
l
u
s
t
r
a
t
i
n
g
t
h
e
p
e
r
f
o
r
m
a
n
c
e
t
r
e
n
d
s
o
f
e
a
c
h
m
o
d
e
l
d
u
r
i
n
g
t
h
e
t
r
a
i
n
i
n
g
p
r
o
c
e
s
s
s
t
a
r
t
i
n
g
f
r
o
m
t
h
e
L
e
N
e
t
(
F
i
g
u
r
e
4
(
a
)
)
,
R
e
s
N
e
t
(
F
i
g
u
r
e
4
(
b
)
)
,
G
o
o
g
l
e
N
e
t
(
F
i
g
u
r
e
4
(
c
)
)
,
V
G
G
N
e
t
(
F
i
g
u
r
e
4
(
d
)
)
,
a
n
d
p
r
o
p
o
s
e
d
m
o
d
e
l
(
R
e
s
N
e
t
+
B
a
g
g
i
n
g
)
m
o
d
e
l
s
(
F
i
g
u
r
e
4
(
e
)
)
.
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
4
.
T
r
ain
in
g
ac
c
u
r
ac
y
c
u
r
v
e
of
(
a)
L
e
N
et,
(
b
)
R
esNet,
(
c)
Go
o
g
leNe
t,
(
d
)
VGGN
et,
an
d
(
e)
p
r
o
p
o
s
ed
m
o
d
el
(
R
esNet+Bag
g
in
g
)
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
0
6
-
5
1
1
9
5112
Fig
u
r
e
4
p
r
esen
ts
th
e
tr
ain
i
n
g
r
esu
lts
,
wh
er
e
th
e
p
r
o
p
o
s
ed
R
esNet+Bag
g
in
g
m
o
d
el
d
em
o
n
s
tr
ates
th
e
b
est
p
er
f
o
r
m
an
ce
am
o
n
g
all
t
h
e
test
ed
m
o
d
els.
T
h
e
tr
ain
in
g
an
d
v
alid
atio
n
ac
c
u
r
ac
y
cu
r
v
es
f
o
r
th
is
m
o
d
el
s
h
o
w
s
ig
n
if
ican
t
s
tab
ilit
y
,
wit
h
th
e
v
alid
atio
n
ac
c
u
r
ac
y
n
ea
r
ly
m
atch
in
g
th
e
tr
ai
n
in
g
ac
cu
r
ac
y
.
T
h
is
in
d
icate
s
th
at
th
e
en
s
em
b
le
lear
n
in
g
m
eth
o
d
with
B
ag
g
in
g
ef
f
ec
t
iv
ely
im
p
r
o
v
ed
R
esNet’
s
g
en
er
aliza
tio
n
ab
ilit
y
,
m
in
im
izin
g
th
e
r
is
k
o
f
o
v
e
r
f
itti
n
g
.
W
ith
th
is
ap
p
r
o
ac
h
,
R
esNet
lev
er
ag
ed
its
s
tr
en
g
th
in
ca
p
tu
r
in
g
co
m
p
le
x
f
ea
tu
r
es
f
r
o
m
b
r
ain
tu
m
o
r
M
R
I
im
ag
es
wh
ile
r
ed
u
ci
n
g
s
en
s
itiv
ity
to
n
o
is
e
in
th
e
d
ata.
T
h
e
h
ig
h
s
tab
ilit
y
an
d
ac
cu
r
ac
y
ac
h
iev
e
d
co
n
f
ir
m
th
e
s
u
p
er
io
r
ity
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
f
o
r
m
ed
ical
im
ag
e
cl
ass
if
icatio
n
,
wh
er
e
h
ig
h
p
r
ec
is
io
n
is
cr
u
cial.
I
n
c
o
m
p
ar
is
o
n
,
o
th
er
m
o
d
els
lik
e
L
eNe
t,
VGGN
et,
an
d
Go
o
g
L
eNe
t
also
s
h
o
wed
g
o
o
d
r
esu
lts
,
th
o
u
g
h
with
ce
r
t
ain
lim
itatio
n
s
.
L
eNe
t,
d
u
e
to
its
s
im
p
ler
ar
ch
itectu
r
e,
ac
h
i
ev
ed
g
o
o
d
tr
ain
in
g
ac
cu
r
ac
y
b
u
t
s
tr
u
g
g
led
to
r
ea
c
h
o
p
tim
al
v
alid
atio
n
ac
cu
r
ac
y
.
Me
an
wh
ile,
Go
o
g
L
eNe
t
an
d
VGGN
et
ex
h
ib
ited
m
o
r
e
s
tab
le
p
er
f
o
r
m
a
n
ce
c
o
m
p
ar
ed
to
L
eNe
t,
b
u
t
th
eir
v
ali
d
atio
n
ac
c
u
r
ac
y
f
lu
ct
u
atio
n
s
,
t
h
o
u
g
h
s
m
aller
th
an
L
eNe
t’
s
,
wer
e
s
til
l
m
o
r
e
p
r
o
n
o
u
n
ce
d
th
a
n
th
o
s
e
o
f
R
esNet+
B
ag
g
in
g
,
in
d
icatin
g
s
o
m
e
lim
i
tatio
n
s
in
h
an
d
lin
g
d
ata
v
ar
iatio
n
s
.
T
h
er
ef
o
r
e,
b
y
co
m
b
in
in
g
th
e
s
tr
en
g
th
s
o
f
R
esNet
wi
th
th
e
s
tab
ilit
y
p
r
o
v
id
ed
b
y
B
ag
g
in
g
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
em
e
r
g
es
as
t
h
e
b
est
s
o
lu
tio
n
f
o
r
b
r
ain
tu
m
o
r
MRI
class
if
icatio
n
.
L
o
s
s
es
in
ea
ch
tr
ain
in
g
ca
n
b
e
s
ee
n
in
Fig
u
r
e
5
,
wh
er
e
F
ig
u
r
es
5
(
a)
to
5
(
d
)
s
h
o
w
th
e
lo
s
s
cu
r
v
es
o
f
L
eNe
t,
R
esNet
,
Go
o
g
L
eNe
t,
an
d
VGGN
et,
r
esp
ec
tiv
ely
,
an
d
Fig
u
r
e
5
(
e)
p
r
esen
ts
th
e
p
r
o
p
o
s
e
d
R
esNet+Bag
g
in
g
m
o
d
el.
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
5
.
T
r
ain
in
g
l
o
s
s
cu
r
v
e
of
(
a)
L
eNe
t,
(
b
)
R
esNet,
(
c)
Go
o
g
leNe
t,
(
d
)
VGGN
et,
an
d
(
e)
p
r
o
p
o
s
ed
m
o
d
el
(
R
esNet+
B
ag
g
in
g
)
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
Op
timiz
in
g
b
r
a
in
tu
mo
r
MRI
cla
s
s
ifica
tio
n
u
s
in
g
a
d
va
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
… (
A
kim
Ma
n
a
o
r
Ha
r
a
P
a
r
d
ed
e
)
5113
Fig
u
r
e
5
illu
s
tr
ates
th
e
tr
ain
in
g
an
d
v
alid
atio
n
lo
s
s
cu
r
v
es,
h
ig
h
lig
h
tin
g
th
e
p
er
f
o
r
m
a
n
ce
d
if
f
er
en
ce
s
b
etwe
en
th
e
L
eNe
t,
R
esNet,
Go
o
g
L
eNe
t,
VGGN
et,
an
d
th
e
p
r
o
p
o
s
ed
m
o
d
el.
T
h
e
lo
s
s
g
r
ap
h
s
f
o
r
all
f
iv
e
m
o
d
els
s
h
o
w
v
ar
y
in
g
p
atter
n
s
in
r
ed
u
cin
g
lo
s
s
d
u
r
in
g
tr
ai
n
in
g
an
d
v
alid
atio
n
.
L
eNe
t
ex
h
ib
its
f
lu
ctu
atin
g
an
d
u
n
s
tab
le
v
alid
atio
n
lo
s
s
,
in
d
icatin
g
th
e
d
if
f
icu
lty
th
e
m
o
d
el
h
as
in
lear
n
in
g
co
m
p
le
x
p
atter
n
s
f
r
o
m
th
e
d
ata
.
R
esNet
d
em
o
n
s
tr
ates
a
co
n
s
is
ten
t
d
ec
lin
e
i
n
lo
s
s
f
o
r
b
o
th
t
r
ain
in
g
a
n
d
v
alid
atio
n
,
alth
o
u
g
h
th
e
r
e
is
a
s
lig
h
t
f
lu
ctu
atio
n
in
t
h
e
v
ali
d
atio
n
lo
s
s
,
s
u
g
g
esti
n
g
th
e
m
o
d
el
m
ay
n
ee
d
a
d
d
itio
n
al
r
eg
u
lar
i
za
tio
n
.
Go
o
g
L
eNe
t
p
er
f
o
r
m
s
well,
with
a
s
tab
le
r
ed
u
ctio
n
in
l
o
s
s
,
th
o
u
g
h
its
v
alid
atio
n
lo
s
s
is
s
lig
h
tly
h
ig
h
er
th
an
th
e
tr
ai
n
in
g
lo
s
s
.
VGGN
et
s
h
o
ws
a
s
ig
n
if
ican
t
d
r
o
p
i
n
tr
ain
in
g
l
o
s
s
,
b
u
t
its
v
alid
atio
n
lo
s
s
r
em
ain
s
u
n
s
tab
le,
h
in
tin
g
at
m
ild
o
v
er
f
itti
n
g
.
I
n
co
n
tr
ast,
th
e
p
r
o
p
o
s
ed
R
esNet+
B
ag
g
in
g
m
o
d
el
ex
h
ib
its
a
v
er
y
s
tab
le
lo
s
s
tr
en
d
,
with
v
alid
atio
n
lo
s
s
lo
wer
th
an
th
at
o
f
th
e
o
th
er
m
o
d
els.
T
h
is
in
d
icate
s
th
at
th
e
B
ag
g
in
g
m
eth
o
d
s
u
cc
ess
f
u
lly
en
h
an
ce
d
t
h
e
m
o
d
el'
s
g
en
er
al
izatio
n
an
d
s
tab
ilit
y
.
As
a
r
esu
lt,
th
e
p
r
o
p
o
s
ed
m
o
d
el
p
r
o
v
es
to
b
e
th
e
m
o
s
t
ef
f
ec
tiv
e
in
h
an
d
lin
g
co
m
p
lex
d
ata,
s
u
ch
as
b
r
ain
tu
m
o
r
M
R
I
im
ag
es.
Fig
u
r
e
6
s
h
o
ws
th
e
R
OC
cu
r
v
e,
wh
er
e
Fig
u
r
es
6
(
a)
to
6
(
d
)
s
h
o
w
th
e
R
OC
cu
r
v
es
o
f
L
eNe
t,
R
esNet,
Go
o
g
L
eNe
t,
an
d
VGGN
et,
r
esp
ec
tiv
ely
,
an
d
Fig
u
r
e
6
(
e)
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
R
esNet+
B
ag
g
in
g
m
o
d
el.
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
6
.
R
OC
cu
r
v
e
of
(
a)
L
e
N
et,
(
b
)
R
esNet,
(
c)
Go
o
g
leNe
t,
(
d
)
VGGN
et,
an
d
(
e)
p
r
o
p
o
s
ed
m
o
d
el
(
R
esNet+
B
ag
g
in
g
)
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
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
5
1
0
6
-
5
1
1
9
5114
Fig
u
r
e
6
s
h
o
ws
th
e
R
OC
cu
r
v
e
f
o
r
all
f
iv
e
m
o
d
els,
r
ef
lectin
g
th
eir
o
v
e
r
all
p
er
f
o
r
m
an
ce
in
d
is
tin
g
u
is
h
in
g
b
etwe
en
b
r
ain
t
u
m
o
r
class
es.
L
eNe
t
ex
h
ib
its
a
R
OC
cu
r
v
e
th
at
clo
s
ely
alig
n
s
with
th
e
d
iag
o
n
al
lin
e,
in
d
icatin
g
its
lim
ited
cla
s
s
if
icatio
n
ab
ilit
y
,
with
a
l
o
w
ar
ea
AUC.
R
esNet
p
r
esen
ts
a
h
ig
h
er
R
OC
cu
r
v
e
co
m
p
ar
ed
to
L
eNe
t,
d
e
m
o
n
s
tr
atin
g
its
b
etter
ab
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
class
es.
Go
o
g
L
eNe
t
also
s
h
o
ws
a
d
ec
en
t
R
OC
cu
r
v
e,
with
an
A
UC
s
im
ilar
to
R
es
Net,
b
u
t
it
is
s
lig
h
tly
less
o
p
tim
al
in
s
o
m
e
class
if
icatio
n
ar
ea
s
.
VGGN
et
'
s
R
OC
p
er
f
o
r
m
an
ce
is
co
m
p
ar
ab
le
to
G
o
o
g
L
eNe
t'
s
,
th
o
u
g
h
it
is
s
o
m
ew
h
at
less
s
tab
le.
I
n
c
o
n
tr
ast,
th
e
p
r
o
p
o
s
ed
R
esNet+Bag
g
in
g
m
o
d
el
s
h
o
ws
th
e
h
ig
h
est
an
d
m
o
s
t
s
tab
le
R
OC
cu
r
v
e
a
m
o
n
g
all
th
e
m
o
d
els,
in
d
icatin
g
s
u
p
e
r
io
r
class
if
icatio
n
ab
ilit
y
with
an
AUC
n
ea
r
i
n
g
th
e
m
ax
im
u
m
v
al
u
e.
T
h
is
h
ig
h
lig
h
ts
th
at
t
h
e
u
s
e
o
f
B
ag
g
in
g
with
R
esNet
s
ig
n
if
ican
tly
en
h
an
ce
s
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
h
e
im
a
g
e
d
is
p
lay
ed
p
r
o
v
id
es
an
ex
am
p
le
o
f
class
if
icatio
n
r
esu
lts
f
r
o
m
th
e
p
r
o
p
o
s
ed
C
NN
m
o
d
el
f
o
r
d
etec
tin
g
b
r
ai
n
tu
m
o
r
MRI
s
ca
n
s
.
A
d
etailed
ex
p
lan
atio
n
o
f
t
h
e
class
if
icatio
n
r
esu
lts
s
h
o
wn
in
Fig
u
r
e
7
.
Fig
u
r
e
7
d
is
p
lay
s
th
e
class
if
icatio
n
r
esu
lts
o
f
th
e
p
r
o
p
o
s
ed
C
NN
m
o
d
el,
s
h
o
wca
s
in
g
ex
am
p
le
p
r
ed
ictio
n
s
o
n
a
b
r
ain
tu
m
o
r
MRI
d
ataset.
Nin
e
im
a
g
es
ar
e
ar
r
an
g
e
d
in
a
3
×
3
m
atr
ix
,
with
ea
ch
im
ag
e
lab
eled
to
s
h
o
w
b
o
th
th
e
tr
u
e
class
(
tr
u
e)
an
d
th
e
p
r
ed
icted
class
(
p
r
ed
)
f
o
r
th
e
ca
teg
o
r
ies
o
f
m
ali
g
n
an
t
(
ca
n
ce
r
)
a
n
d
n
o
r
m
al
(
n
o
ca
n
ce
r
)
.
Sev
er
al
p
r
ed
ictio
n
s
ar
e
c
o
r
r
ec
t,
s
u
ch
as
th
e
Ma
lig
n
an
t
ca
s
es
co
r
r
ec
tly
p
r
ed
icted
as
Ma
lig
n
an
t,
an
d
th
e
N
o
r
m
al
ca
s
es
ac
cu
r
ately
p
r
e
d
icted
as
n
o
r
m
al.
Ho
wev
er
,
th
er
e
ar
e
also
s
o
m
e
m
is
class
if
icatio
n
s
,
s
u
ch
as
a
n
o
r
m
al
ca
s
e
in
co
r
r
ec
tly
p
r
ed
ict
ed
as
Ma
lig
n
an
t.
T
h
e
p
ix
elate
d
ap
p
ea
r
an
ce
o
f
th
e
im
ag
es
r
ef
lects
th
e
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
ap
p
lied
to
e
n
h
an
ce
th
e
m
o
d
el'
s
ab
ilit
y
to
d
etec
t
k
ey
v
is
u
al
p
atter
n
s
.
W
h
ile
th
e
m
o
d
el
p
er
f
o
r
m
s
w
ell
in
id
en
tify
in
g
m
o
s
t
p
atter
n
s
,
th
e
m
is
class
if
icatio
n
s
h
ig
h
lig
h
t
th
e
n
ee
d
f
o
r
f
u
r
th
er
r
e
f
in
em
en
t,
s
u
ch
as
a
d
ju
s
tm
en
ts
to
p
ar
am
eter
s
elec
tio
n
o
r
ar
ch
itectu
r
al
s
tr
u
ctu
r
e,
to
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
an
d
r
eliab
ilit
y
o
f
b
r
ai
n
tu
m
o
r
MRI
class
if
icatio
n
.
Fig
u
r
e
7
.
C
lass
if
icatio
n
r
esu
lts
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
3
.
3
.
Dis
cus
s
io
n
T
h
e
tr
ain
in
g
an
d
test
in
g
r
esu
lt
s
in
clu
d
e
co
n
f
u
s
io
n
m
atr
ices
f
o
r
ea
ch
C
NN
m
o
d
el,
h
ig
h
lig
h
tin
g
th
eir
class
if
icatio
n
p
er
f
o
r
m
an
ce
ac
r
o
s
s
th
r
ee
ca
teg
o
r
ies:
b
en
ig
n
,
m
alig
n
an
t,
an
d
n
o
r
m
al
b
r
a
in
tu
m
o
r
s
.
T
h
ese
m
atr
ices
p
r
o
v
id
e
a
clea
r
v
ie
w
o
f
ea
ch
m
o
d
el’
s
s
tr
en
g
th
s
an
d
wea
k
n
ess
es
in
d
is
tin
g
u
is
h
in
g
b
etwe
en
th
e
class
es.
Fig
u
r
e
8
illu
s
tr
ates
th
e
co
n
f
u
s
io
n
m
atr
ices,
allo
win
g
f
o
r
d
ir
ec
t
co
m
p
a
r
is
o
n
o
f
cla
s
s
if
icatio
n
ac
cu
r
ac
y
an
d
er
r
o
r
d
is
tr
ib
u
tio
n
am
o
n
g
t
h
e
ev
alu
ated
m
o
d
els.
Fig
u
r
e
8
illu
s
tr
ates
th
e
co
n
f
u
s
io
n
m
atr
ices
f
o
r
th
e
f
iv
e
m
o
d
els,
s
h
o
win
g
v
ar
y
i
n
g
p
er
f
o
r
m
an
ce
in
class
if
y
in
g
h
em
o
r
r
h
a
g
e
an
d
n
o
n
-
h
em
o
r
r
h
ag
e
d
ata.
L
eNe
t
as
s
h
o
wn
in
Fig
u
r
e
8
(
a)
d
em
o
n
s
tr
ates
m
o
d
er
ate
p
er
f
o
r
m
an
ce
,
with
s
ev
er
al
p
r
e
d
ictio
n
er
r
o
r
s
in
b
o
th
class
es.
R
esNet
as
s
h
o
wn
in
Fig
u
r
e
8
(
b
)
p
er
f
o
r
m
s
s
lig
h
tly
b
etter
th
an
L
eNe
t,
co
r
r
ec
tly
p
r
ed
ictin
g
m
o
r
e
in
s
tan
ce
s
o
f
th
e
h
em
o
r
r
h
ag
e
class
.
Go
o
g
leNe
t
as
s
h
o
wn
in
Fig
u
r
e
8
(
c)
s
h
o
ws s
o
m
e
im
b
alan
ce
in
its
p
r
ed
ictio
n
s
,
p
er
f
o
r
m
in
g
less
o
p
tim
ally
f
o
r
th
e
n
o
n
-
h
em
o
r
r
h
a
g
e
class
.
VGGN
et
as
s
h
o
wn
in
Fig
u
r
e
8
(
d
)
s
h
o
ws
an
im
p
r
o
v
e
m
en
t
i
n
ac
cu
r
ac
y
o
v
e
r
Go
o
g
leNe
t,
with
a
m
o
r
e
b
alan
ce
d
p
r
ed
ictio
n
d
is
tr
ib
u
tio
n
ac
r
o
s
s
b
o
th
class
es.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
(
R
esNet+Bag
g
in
g
)
as
s
h
o
wn
in
Fig
u
r
e
8
(
e)
d
eliv
er
s
th
e
b
est p
er
f
o
r
m
an
ce
,
with
th
e
f
ewe
s
t p
r
ed
ictio
n
er
r
o
r
s
am
o
n
g
all
m
o
d
els,
r
ef
lecti
n
g
th
e
ef
f
ec
tiv
en
ess
o
f
en
s
em
b
le
lear
n
in
g
in
en
h
a
n
cin
g
class
if
icatio
n
ac
cu
r
ac
y
.
T
h
e
r
esu
lts
ar
e
p
r
esen
ted
i
n
T
ab
le
2
.
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
Op
timiz
in
g
b
r
a
in
tu
mo
r
MRI
cla
s
s
ifica
tio
n
u
s
in
g
a
d
va
n
ce
d
p
r
ep
r
o
ce
s
s
in
g
… (
A
kim
Ma
n
a
o
r
Ha
r
a
P
a
r
d
ed
e
)
5115
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
8
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
(
a)
L
e
N
et,
(
b
)
R
esNet,
(
c)
Go
o
g
leNe
t,
(
d
)
VGGN
et,
an
d
(
e)
p
r
o
p
o
s
ed
mo
d
el
(
R
esNet+
B
ag
g
in
g
)
T
ab
le
2
s
h
o
ws
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
(
R
esNet+Bag
g
in
g
)
o
u
tp
er
f
o
r
m
s
all
o
th
er
m
o
d
els
test
ed
.
W
ith
a
test
in
g
ac
cu
r
ac
y
o
f
7
2
%,
th
is
m
o
d
el
d
em
o
n
s
tr
ates
s
ig
n
if
ican
t
g
en
er
aliza
tio
n
ca
p
ab
ilit
ies,
f
ar
s
u
r
p
ass
in
g
o
th
er
m
o
d
els
s
u
c
h
as
L
eNe
t
(
4
5
%),
s
tan
d
a
r
d
R
esNet
(
5
7
%),
Go
o
g
leNe
t
(
4
5
%),
an
d
VGGN
et
(
5
0
%).
Fu
r
th
er
m
o
r
e,
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
also
r
ec
o
r
d
s
th
e
h
ig
h
est
p
r
ec
is
io
n
(
0
.
6
0
1
0
)
,
r
e
ca
ll
(
0
.
6
0
0
0
)
,
an
d
F1
-
s
co
r
e
(
0
.
5
9
9
0
)
am
o
n
g
t
h
e
o
th
er
m
o
d
els,
in
d
icatin
g
a
g
o
o
d
b
alan
ce
b
etwe
en
d
etec
tin
g
p
o
s
itiv
e
an
d
n
eg
ativ
e
class
es.
I
ts
s
p
ec
if
icit
y
(
0
.
5
5
0
0
)
an
d
MCC
(
0
.
2
0
1
0
)
also
s
h
o
w
an
ad
v
a
n
tag
e
in
h
a
n
d
lin
g
im
b
alan
ce
d
Evaluation Warning : The document was created with Spire.PDF for Python.