I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
,
p
p
.
2809
~
2
8
1
9
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
3
.
pp
2
8
0
9
-
2
8
1
9
2809
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Im
pro
v
ing
breas
t
ca
ncer c
la
ss
ifi
ca
t
io
n wit
h a
no
v
el
VG
G
1
9
-
ba
sed en
sem
ble learning
a
ppro
a
ch
Cha
y
m
a
e
T
a
ib,
Adna
n E
l A
hm
a
di,
O
t
ma
n Abdo
un
,
E
l
K
ha
t
ir
H
a
imo
ud
i
I
n
f
o
r
mat
i
o
n
S
e
c
u
r
i
t
y
,
I
n
t
e
l
l
i
g
e
n
t
S
y
s
t
e
ms a
n
d
A
p
p
l
i
c
a
t
i
o
n
,
F
a
c
u
l
t
y
o
f
sc
i
e
n
c
e
T
e
t
o
u
a
n
,
A
b
d
e
l
mal
e
k
Ess
a
a
d
i
U
n
i
v
e
r
si
t
y
,
Te
t
o
u
a
n
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
1
4
,
2
0
2
4
R
ev
is
ed
J
an
2
7
,
2
0
2
5
Acc
ep
ted
Ma
r
3
,
2
0
2
5
Bre
a
st
c
a
n
c
e
r
is
o
n
e
o
f
th
e
m
o
st
li
fe
-
th
re
a
ten
in
g
d
ise
a
se
s,
p
a
rti
c
u
larly
a
ffe
c
ti
n
g
wo
m
e
n
,
h
i
g
h
li
g
h
ti
n
g
t
h
e
imp
o
rtan
c
e
o
f
e
a
rly
d
e
tec
ti
o
n
fo
r
i
m
p
ro
v
i
n
g
su
rv
iv
a
l
ra
tes
.
In
t
h
is
stu
d
y
,
we
p
ro
p
o
se
a
n
o
v
e
l
d
iag
n
o
stic
fra
m
e
wo
rk
th
a
t
c
o
m
b
in
e
s
a
m
o
d
ifi
e
d
VG
G
1
9
a
rc
h
it
e
c
tu
re
wit
h
Ba
g
g
in
g
e
n
se
m
b
le
lea
rn
in
g
,
u
sin
g
th
re
e
b
a
se
c
las
sifiers
:
d
e
c
is
io
n
tree
(DT),
lo
g
isti
c
re
g
re
ss
io
n
(LR)
,
a
n
d
su
p
p
o
r
t
v
e
c
to
r
m
a
c
h
in
e
(S
V
M
).
We
a
lso
c
o
m
p
a
re
th
is
a
p
p
r
o
a
c
h
with
twe
n
ty
-
fo
u
r
h
y
b
rid
m
o
d
e
ls,
i
n
teg
r
a
ti
n
g
v
a
rio
u
s
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
(CNN
)
a
rc
h
it
e
c
tu
re
s
(Re
sN
e
t5
0
,
VG
G
1
9
,
Co
n
v
Ne
x
tBas
e
,
De
n
se
Ne
t1
2
1
,
Eff
icie
n
tNe
tV2
B0
,
Eff
icie
n
tNe
t
B0
,
M
o
b
i
leN
e
t,
a
n
d
Na
sN
e
tM
o
b
il
e
)
wit
h
Ba
g
g
in
g
e
n
se
m
b
le
m
e
th
o
d
s.
O
u
r
re
su
lt
s
sh
o
w
th
a
t
t
h
e
p
r
o
p
o
se
d
m
o
d
e
l
o
u
t
p
e
rfo
rm
s
a
ll
o
th
e
r
a
rc
h
it
e
c
t
u
r
e
s,
e
sp
e
c
ially
wh
e
n
c
o
m
b
in
e
d
w
it
h
S
V
M
,
a
c
h
iev
in
g
a
c
c
u
ra
c
y
o
f
9
7
%
o
n
t
h
e
f
in
e
n
e
e
d
le
a
sp
iratio
n
c
y
t
o
lo
g
y
(F
NA
C)
d
a
tas
e
t
a
n
d
9
0
%
o
n
th
e
In
ter
n
a
ti
o
n
a
l
Co
n
fe
re
n
c
e
o
n
Im
a
g
e
An
a
ly
sis
a
n
d
Re
c
o
g
n
it
i
o
n
(
ICIAR
)
d
a
tas
e
t.
Th
i
s
fra
m
e
wo
rk
d
e
m
o
n
stra
tes
stro
n
g
p
o
ten
ti
a
l
fo
r
imp
r
o
v
i
n
g
e
a
rly
b
re
a
st ca
n
c
e
r
d
iag
n
o
sis.
K
ey
w
o
r
d
s
:
B
ag
g
in
g
B
r
ea
s
t c
an
ce
r
C
las
s
if
icatio
n
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
Dee
p
lear
n
in
g
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
:
C
h
ay
m
ae
T
aib
I
n
f
o
r
m
atio
n
Secu
r
ity
,
I
n
tellig
e
n
t Sy
s
tem
s
an
d
Ap
p
licatio
n
,
F
ac
u
lty
o
f
Scien
ce
T
et
o
u
an
,
Ab
d
elm
alek
E
s
s
aa
d
i U
n
iv
er
s
ity
T
eto
u
an
,
M
o
r
o
cc
o
E
m
ail: ta
ib
.
ch
ay
m
ae
@
etu
.
u
ae
.
ac
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
B
r
ea
s
t
ca
n
ce
r
is
o
n
e
o
f
th
e
m
o
s
t
co
m
m
o
n
an
d
life
-
th
r
ea
ten
i
n
g
ca
n
ce
r
s
af
f
ec
tin
g
wo
m
e
n
wo
r
ld
wid
e.
Acc
o
r
d
in
g
t
o
th
e
W
o
r
ld
Hea
lt
h
Or
g
an
izatio
n
(
W
HO)
[
1
]
,
it
a
cc
o
u
n
ts
f
o
r
ap
p
r
o
x
im
ately
2
5
%
o
f
all
ca
n
ce
r
ca
s
es
am
o
n
g
wo
m
en
,
m
ak
i
n
g
it
th
e
m
o
s
t
p
r
e
v
alen
t
m
alig
n
an
c
y
i
n
th
is
p
o
p
u
latio
n
.
T
h
e
in
cid
e
n
ce
o
f
b
r
ea
s
t
ca
n
ce
r
v
ar
ies ac
r
o
s
s
r
eg
io
n
s
,
with
h
ig
h
er
r
ates o
b
s
er
v
ed
in
d
ev
el
o
p
e
d
co
u
n
tr
ies.
T
h
is
v
ar
iatio
n
is
lik
e
ly
in
f
lu
e
n
ce
d
b
y
f
ac
to
r
s
s
u
ch
as
life
s
ty
le
ch
o
ic
es,
r
ep
r
o
d
u
ctiv
e
b
e
h
av
io
r
s
,
a
n
d
m
o
r
e
a
d
v
an
ce
d
s
cr
ee
n
in
g
p
r
o
g
r
am
s
.
Acc
u
r
ate
d
iag
n
o
s
is
is
ess
en
tial
f
o
r
d
eter
m
in
in
g
th
e
m
o
s
t
ef
f
ec
tiv
e
tr
ea
tm
en
t
p
lan
[
2
]
.
Ad
v
an
c
es
in
im
ag
in
g
tech
n
iq
u
es
lik
e
m
am
m
o
g
r
a
p
h
y
,
u
ltra
s
o
u
n
d
,
an
d
m
a
g
n
etic
r
eso
n
an
ce
im
ag
in
g
(
MRI)
h
av
e
s
ig
n
if
ic
an
tly
im
p
r
o
v
e
d
th
e
p
r
ec
is
io
n
o
f
d
etec
tin
g
b
r
ea
s
t
ca
n
ce
r
lesi
o
n
s
.
Ad
d
itio
n
ally
,
m
o
lecu
lar
an
d
g
en
etic
test
in
g
o
f
tu
m
o
r
s
am
p
les
p
r
o
v
id
es
cr
itical
in
s
ig
h
ts
in
to
th
e
ch
ar
ac
ter
is
tics
o
f
th
e
ca
n
ce
r
,
en
ab
lin
g
p
er
s
o
n
alize
d
t
r
ea
tm
en
ts
th
at
ca
n
en
h
an
ce
p
atien
t o
u
tco
m
es
[
3
]
.
Ma
ch
in
e
lear
n
in
g
(
ML
)
tech
n
i
q
u
es
h
av
e
b
r
o
u
g
h
t
a
r
ev
o
lu
tio
n
in
m
ed
ical
d
iag
n
o
s
is
,
o
f
f
er
in
g
p
o
wer
f
u
l
to
o
ls
f
o
r
an
aly
zin
g
co
m
p
lex
d
atasets
an
d
id
en
tify
in
g
p
atte
r
n
s
th
at
m
ig
h
t
b
e
m
is
s
ed
b
y
tr
ad
itio
n
al
s
tatis
tical
m
eth
o
d
s
[
4
]
.
ML
en
co
m
p
ass
es
v
ar
io
u
s
alg
o
r
ith
m
s
an
d
m
o
d
el
s
,
in
clu
d
in
g
s
u
p
er
v
is
ed
[
5
]
–
[
7
]
,
u
n
s
u
p
er
v
is
ed
[
8
]
,
an
d
r
ei
n
f
o
r
ce
m
en
t
lear
n
in
g
[
9
]
,
[
1
0
]
,
as
well
as
d
ee
p
lear
n
i
n
g
tech
n
iq
u
es
[
1
1
]
.
Fo
r
b
r
ea
s
t
ca
n
ce
r
d
iag
n
o
s
is
,
ea
r
ly
an
d
ac
c
u
r
ate
id
en
tific
ati
o
n
is
cr
u
cial
f
o
r
im
p
r
o
v
in
g
s
u
r
v
iv
al
r
ates
[
1
2
]
,
[
1
3
]
.
W
h
ile
a
v
ar
iety
o
f
im
a
g
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
8
0
9
-
2
8
1
9
2810
tech
n
iq
u
es,
s
u
ch
as
MRI,
m
a
m
m
o
g
r
a
p
h
y
,
u
ltra
s
o
u
n
d
,
an
d
t
h
er
m
o
g
r
ap
h
y
,
ar
e
u
s
ed
f
o
r
d
et
ec
tio
n
,
ea
ch
m
et
h
o
d
h
as its
o
wn
ad
v
an
tag
es a
n
d
li
m
itatio
n
s
.
R
ec
en
tly
,
h
y
b
r
id
m
o
d
els
co
m
b
in
in
g
m
ac
h
in
e
lea
r
n
in
g
an
d
d
ee
p
lear
n
i
n
g
h
av
e
g
ain
ed
atten
tio
n
as
a
way
to
im
p
r
o
v
e
b
r
ea
s
t c
an
ce
r
d
iag
n
o
s
is
.
Fo
r
ex
am
p
le
,
Kh
a
n
et
a
l
.
d
ev
elo
p
e
d
a
d
ee
p
lear
n
in
g
f
r
a
m
ewo
r
k
u
s
in
g
tr
an
s
f
er
lear
n
in
g
f
o
r
b
r
ea
s
t
ca
n
ce
r
class
if
icatio
n
in
cy
to
lo
g
y
im
ag
es,
ac
h
iev
in
g
9
7
.
5
2
%
ac
cu
r
ac
y
s
u
r
p
ass
in
g
ex
is
tin
g
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN)
m
o
d
els
lik
e
Go
o
g
L
eNe
t,
VGGN
et,
an
d
R
esNet
[
1
4
]
.
Atb
an
et
al
.
p
r
o
p
o
s
ed
a
f
u
s
ed
m
o
d
el
co
m
b
in
in
g
tr
an
s
f
er
lear
n
in
g
an
d
m
etah
eu
r
is
tic
alg
o
r
ith
m
s
f
o
r
h
is
to
p
ath
o
lo
g
ical
im
ag
e
class
if
icatio
n
,
ac
h
iev
i
n
g
9
7
.
7
3
%
ac
c
u
r
ac
y
an
d
a
9
7
.
7
5
%
F1
-
s
co
r
e
[
1
5
]
.
Similar
ly
,
Qasr
awi
et
a
l
.
d
esig
n
ed
a
h
y
b
r
id
d
ee
p
lear
n
in
g
en
s
em
b
le
m
o
d
el
f
o
r
m
a
m
m
o
g
r
am
a
n
aly
s
is
,
r
ea
ch
in
g
9
9
.
7
%
ac
cu
r
ac
y
f
o
r
b
en
ig
n
ca
s
es
an
d
9
9
.
8
%
f
o
r
m
alig
n
an
t
o
n
es
[
1
6
]
.
Oth
er
ap
p
r
o
ac
h
es,
s
u
ch
as
R
aa
j’
s
h
y
b
r
id
C
NN
ar
ch
itectu
r
e
an
d
Asl
an
’
s
en
d
-
to
-
en
d
C
NN
an
d
C
NN
-
b
ased
b
i
-
d
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
C
NN+
B
iLST
M
)
m
o
d
el,
d
em
o
n
s
tr
ated
im
p
r
ess
iv
e
ac
cu
r
ac
y
r
ates
f
o
r
b
r
ea
s
t
ca
n
ce
r
d
etec
tio
n
in
th
e
m
am
m
o
g
r
a
p
h
ic
im
ag
e
a
n
aly
s
is
s
o
ciety
(
MI
AS)
d
ataset
[
1
7
]
,
[
1
8
]
.
Fu
r
th
er
m
o
r
e,
Sre
ep
r
a
d
a
et
a
l
.
p
r
o
p
o
s
ed
a
h
y
b
r
id
c
o
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
-
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
(
C
NN
-
SVM
)
ap
p
r
o
ac
h
th
at
o
u
tp
e
r
f
o
r
m
ed
tr
a
d
itio
n
al
C
N
N
tech
n
iq
u
es
with
a
9
8
.
7
0
%
ac
cu
r
ac
y
in
lu
n
g
C
T
im
ag
e
class
if
icatio
n
,
d
em
o
n
s
tr
atin
g
th
e
b
r
o
ad
e
r
ap
p
licab
ilit
y
o
f
h
y
b
r
id
m
o
d
els
in
m
ed
ical
im
ag
in
g
[
1
9
]
.
I
n
s
u
m
m
ar
y
,
c
o
m
b
in
in
g
C
NNs
with
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
tech
n
i
q
u
es
an
d
en
s
em
b
l
e
m
eth
o
d
s
lik
e
b
ag
g
in
g
h
as
led
to
s
ig
n
if
ican
t
im
p
r
o
v
em
en
ts
in
b
r
ea
s
t
ca
n
ce
r
d
iag
n
o
s
is
.
T
h
ese
ad
v
an
ce
m
en
ts
r
esu
lt
in
m
o
r
e
ac
cu
r
ate,
r
eliab
le,
a
n
d
r
o
b
u
s
t p
r
ed
ictiv
e
m
o
d
els,
p
a
v
in
g
th
e
way
f
o
r
b
etter
clin
ical
o
u
tc
o
m
es.
T
h
e
p
r
im
ar
y
g
o
al
o
f
th
is
s
tu
d
y
is
to
ev
alu
ate
th
e
ef
f
ec
tiv
en
ess
o
f
co
m
b
in
in
g
C
NNs
an
d
b
ag
g
in
g
tech
n
iq
u
es
to
en
h
a
n
ce
th
e
ac
c
u
r
ac
y
a
n
d
r
eliab
ilit
y
o
f
b
r
ea
s
t
ca
n
ce
r
d
iag
n
o
s
is
.
B
y
lev
er
ag
in
g
C
NNs
'
f
ea
tu
r
e
ex
tr
ac
tio
n
ca
p
ab
ilit
ies
an
d
th
e
en
s
em
b
le
s
tab
ilit
y
o
f
b
a
g
g
in
g
,
th
e
s
tu
d
y
aim
s
t
o
d
e
v
elo
p
a
r
o
b
u
s
t
d
iag
n
o
s
tic
m
o
d
el
th
at
o
u
tp
e
r
f
o
r
m
s
tr
ad
itio
n
al
m
eth
o
d
s
.
I
t
s
ee
k
s
to
ass
es
s
th
e
p
er
f
o
r
m
an
ce
o
f
C
NNs
as
f
ea
tu
r
e
e
x
tr
ac
to
r
s
,
ev
alu
ate
th
e
im
p
ac
t
o
f
b
ag
g
i
n
g
o
n
r
e
d
u
cin
g
v
ar
ia
n
ce
an
d
im
p
r
o
v
in
g
ac
cu
r
ac
y
in
C
NN
-
b
ased
m
o
d
e
ls
,
an
d
co
m
p
ar
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
co
m
b
in
ed
C
NN
-
b
ag
g
i
n
g
ap
p
r
o
ac
h
with
s
tan
d
alo
n
e
C
NNs
an
d
tr
ad
itio
n
al
d
iag
n
o
s
tic
tech
n
iq
u
es.
A
d
d
itio
n
ally
,
t
h
e
s
tu
d
y
e
x
p
lo
r
es
th
e
p
o
ten
tial
o
f
th
e
co
m
b
i
n
ed
m
eth
o
d
i
n
m
i
n
im
izin
g
f
alse
p
o
s
itiv
es
an
d
n
eg
ativ
es
in
b
r
ea
s
t
ca
n
ce
r
d
iag
n
o
s
is
an
d
ev
alu
ates
th
e
m
o
d
if
ied
VG
G1
9
co
m
b
in
ed
with
b
ag
g
in
g
ap
p
r
o
ac
h
a
g
ain
s
t o
th
e
r
tr
ad
itio
n
al
C
NN
tech
n
iq
u
es.
T
h
e
p
a
p
er
is
o
r
g
an
ized
in
to
f
i
v
e
s
ec
tio
n
s
to
e
n
s
u
r
e
clar
ity
a
n
d
co
h
er
en
ce
.
Sectio
n
2
r
e
v
ie
ws
r
elate
d
wo
r
k
,
s
ec
tio
n
3
e
x
p
lain
s
th
e
m
eth
o
d
o
l
o
g
y
,
a
n
d
s
ec
tio
n
4
p
r
esen
ts
th
e
r
esu
lts
.
Sectio
n
5
co
n
clu
d
es
with
k
ey
in
s
ig
h
ts
an
d
s
u
g
g
esti
o
n
s
f
o
r
f
u
tu
r
e
r
esear
ch
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
R
ec
en
t
ad
v
an
ce
m
e
n
ts
in
d
ee
p
lear
n
in
g
an
d
ar
tific
ial
in
tellig
en
ce
h
av
e
s
ig
n
if
ican
tly
en
h
a
n
ce
d
ca
n
ce
r
d
iag
n
o
s
tics
,
p
ar
ticu
lar
ly
in
b
r
e
ast
ca
n
ce
r
class
if
icatio
n
.
Nu
m
e
r
o
u
s
s
tu
d
ies
h
av
e
i
n
v
esti
g
ated
t
h
e
u
s
e
o
f
en
s
em
b
le
lear
n
in
g
m
eth
o
d
s
an
d
d
ee
p
C
NN
ar
ch
itectu
r
es
to
b
o
o
s
t
d
iag
n
o
s
tic
ac
cu
r
ac
y
an
d
m
o
d
el
r
eliab
ilit
y
.
T
h
ese
ap
p
r
o
ac
h
es
h
av
e
d
em
o
n
s
tr
ated
s
tr
o
n
g
p
o
te
n
tial
in
h
a
n
d
lin
g
co
m
p
lex
h
is
to
p
ath
o
lo
g
ical
im
ag
es
an
d
im
p
r
o
v
in
g
clin
ical
d
ec
is
io
n
-
m
ak
in
g
.
B
alasu
b
r
am
an
ian
et
a
l.
u
tili
ze
d
en
s
em
b
le
d
ee
p
lear
n
i
n
g
tech
n
iq
u
es
co
m
b
in
in
g
VGG1
6
an
d
R
esNet5
0
ar
ch
itectu
r
es
to
class
if
y
b
r
ea
s
t
ca
n
ce
r
s
u
b
ty
p
es
an
d
i
n
v
asiv
en
ess
u
s
in
g
th
e
b
r
ea
s
t
ca
n
ce
r
h
is
to
lo
g
(
B
AC
H
)
d
ataset.
T
h
ey
in
tr
o
d
u
ce
d
an
i
m
ag
e
p
atch
in
g
tech
n
iq
u
e
f
o
r
h
ig
h
-
r
eso
lu
tio
n
p
r
ep
r
o
ce
s
s
in
g
,
ac
h
iev
in
g
a
p
atc
h
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
5
.
3
1
%
an
d
a
wh
o
le
-
s
lid
e
im
ag
e
c
lass
if
icatio
n
ac
cu
r
ac
y
o
f
9
8
.
4
3
%
[
2
0
]
.
W
h
ile
th
e
s
tu
d
y
d
em
o
n
s
tr
ates
th
e
ef
f
icac
y
o
f
en
s
em
b
le
m
eth
o
d
s
,
it
r
elies
h
ea
v
ily
o
n
ex
is
tin
g
ar
ch
itectu
r
es
with
o
u
t
s
ig
n
if
ican
t
in
n
o
v
atio
n
.
A
d
d
itio
n
ally
,
its
f
o
cu
s
o
n
p
at
ch
-
lev
el
class
if
icatio
n
r
aises
co
n
ce
r
n
s
ab
o
u
t
g
en
er
aliza
b
ilit
y
ac
r
o
s
s
d
atasets
an
d
r
e
al
-
wo
r
ld
ap
p
licatio
n
s
.
T
h
is
s
tu
d
y
u
n
d
e
r
s
co
r
es
th
e
n
e
ed
f
o
r
m
o
r
e
r
o
b
u
s
t
an
d
ad
ap
tab
le
m
o
d
els,
wh
ich
th
is
r
esear
ch
ad
d
r
ess
es
th
r
o
u
g
h
th
e
d
ee
p
er
ar
c
h
itectu
r
e
an
d
en
h
an
ce
d
e
n
s
em
b
le
f
r
am
ewo
r
k
o
f
VGG1
9
.
Na
d
k
a
r
n
i
an
d
No
r
o
n
h
a
p
r
o
p
o
s
ed
an
en
s
em
b
le
o
f
C
NNs
f
o
r
class
if
y
in
g
m
a
m
m
o
g
r
ap
h
y
im
ag
es
u
s
in
g
th
e
MI
AS
a
n
d
ig
i
tal
d
atab
ase
f
o
r
s
cr
ee
n
in
g
m
am
m
o
g
r
ap
h
y
(
DDSM)
d
atasets
[
2
1
]
.
T
h
eir
a
p
p
r
o
ac
h
m
er
g
ed
th
ese
d
atasets
to
a
d
d
r
e
s
s
is
s
u
es
o
f
d
ataset
im
b
alan
ce
an
d
o
v
er
f
itti
n
g
,
ac
h
ie
v
in
g
an
a
cc
u
r
ac
y
o
f
9
5
.
7
%.
W
h
ile
th
eir
en
s
em
b
le
s
tr
ateg
y
im
p
r
o
v
ed
r
o
b
u
s
tn
ess
,
it
la
ck
ed
ar
ch
itectu
r
al
n
o
v
elty
,
a
n
d
th
eir
f
o
c
u
s
o
n
m
am
m
o
g
r
a
m
s
lim
its
its
d
ir
e
ct
ap
p
licab
ilit
y
to
h
is
to
p
ath
o
lo
g
ical
d
atasets
lik
e
B
AC
H
an
d
f
in
e
n
ee
d
le
asp
ir
atio
n
cy
to
lo
g
y
(
FNAC
)
.
Nev
er
th
el
ess
,
th
eir
ef
f
o
r
ts
to
m
itig
ate
o
v
er
f
itti
n
g
an
d
h
an
d
le
d
i
v
er
s
e
d
atasets
p
r
o
v
id
e
v
alu
ab
le
in
s
ig
h
ts
f
o
r
th
is
s
tu
d
y
.
Ma
m
d
y
an
d
Petli
[
2
2
]
d
ev
e
lo
p
ed
a
co
m
p
u
tatio
n
ally
ef
f
icien
t
f
r
am
ewo
r
k
f
o
r
MRI
-
b
ased
b
r
ea
s
t
ca
n
ce
r
d
ete
ctio
n
u
s
in
g
a
m
o
d
if
ie
d
f
u
zz
y
r
o
u
g
h
s
et
tech
n
i
q
u
e
an
d
s
tack
e
d
au
to
e
n
co
d
e
r
s
.
T
h
is
m
eth
o
d
ac
h
iev
e
d
a
r
em
a
r
k
ab
le
9
9
%
class
if
icatio
n
ac
c
u
r
ac
y
wh
ile
m
in
im
izin
g
co
m
p
u
ta
tio
n
al
co
m
p
lex
ity
.
Alth
o
u
g
h
f
o
cu
s
ed
o
n
MRI
im
a
g
in
g
r
ath
e
r
th
an
h
is
to
p
ath
o
lo
g
y
,
th
eir
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
tech
n
iq
u
es
o
f
f
e
r
in
s
ig
h
ts
in
to
o
p
tim
izin
g
f
ea
t
u
r
e
ex
t
r
ac
tio
n
,
wh
ich
c
o
u
ld
co
m
p
lem
en
t
th
is
s
tu
d
y
'
s
en
s
em
b
le
ap
p
r
o
ac
h
b
y
en
h
an
cin
g
th
e
ef
f
icien
cy
o
f
f
e
atu
r
es
d
er
i
v
e
d
f
r
o
m
VGG1
9
.
Mu
d
av
ad
k
ar
et
a
l.
[
2
3
]
ap
p
lied
en
s
em
b
le
lea
r
n
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
I
mp
r
o
vin
g
b
r
ea
s
t c
a
n
ce
r
cla
s
s
ifica
tio
n
w
ith
a
n
o
ve
l V
GG1
9
-
b
a
s
ed
en
s
emb
le
…
(
C
h
a
ym
a
e
Ta
ib
)
2811
to
g
astric
ca
n
ce
r
d
etec
tio
n
u
s
in
g
th
e
GasHis
S
DB
d
ata
s
et.
B
y
co
m
b
in
in
g
R
esNet5
0
,
VGG
Net,
an
d
R
esNet3
4
,
th
ey
ac
h
ie
v
ed
o
v
er
9
9
%
ac
c
u
r
ac
y
ac
r
o
s
s
v
a
r
io
u
s
r
eso
lu
ti
o
n
s
.
Alth
o
u
g
h
u
n
r
elate
d
to
b
r
ea
s
t
ca
n
ce
r
,
th
is
s
tu
d
y
h
ig
h
lig
h
ts
th
e
p
o
ten
tial
o
f
en
s
em
b
le
m
o
d
els
to
ca
p
tu
r
e
cr
iti
ca
l
f
ea
tu
r
es
ac
r
o
s
s
r
eso
lu
tio
n
s
.
Ho
wev
er
,
it
lack
s
s
p
ec
if
icity
r
eg
ar
d
in
g
b
r
ea
s
t
ca
n
ce
r
d
atasets
,
lim
itin
g
its
ap
p
licab
ilit
y
to
th
is
r
e
s
ea
r
ch
'
s
f
o
cu
s
o
n
h
is
to
p
ath
o
lo
g
ical
d
ata.
Ko
n
d
e
jk
ar
et
a
l.
[
2
4
]
.
d
em
o
n
s
tr
ated
th
e
ef
f
ec
tiv
en
ess
o
f
R
esNet
m
o
d
els
in
p
r
o
s
tate
ca
n
ce
r
g
r
a
d
in
g
u
s
in
g
m
u
lti
-
s
c
ale
p
atch
-
lev
el
p
ath
o
lo
g
y
im
a
g
es,
ac
h
iev
in
g
n
ea
r
-
p
er
f
ec
t
ac
cu
r
ac
y
o
f
0
.
9
9
9
i
n
id
en
tify
in
g
clin
ically
s
ig
n
if
ica
n
t
ca
n
ce
r
.
W
h
ile
th
is
s
tu
d
y
em
p
h
asizes
th
e
s
ca
lab
ilit
y
an
d
p
r
ec
is
io
n
o
f
C
NNs,
its
f
o
cu
s
o
n
p
r
o
s
tate
ca
n
ce
r
an
d
g
r
ad
in
g
d
iv
er
g
es
f
r
o
m
th
e
class
if
icatio
n
o
b
jectiv
es o
f
th
is
r
esear
ch
.
No
n
eth
eless
,
its
s
u
cc
es
s
in
p
atch
-
lev
el
an
al
y
s
is
r
ein
f
o
r
ce
s
th
e
p
o
ten
tial
o
f
C
NNs
lik
e
VG
G1
9
f
o
r
h
is
to
p
ath
o
lo
g
ical
task
s
.
Sin
g
h
et
a
l.
[
2
5
]
.
in
tr
o
d
u
ce
d
an
AI
-
b
ased
web
ap
p
licatio
n
in
teg
r
atin
g
E
f
f
icien
tNet
-
B
1
f
o
r
p
r
o
s
tate
ca
n
ce
r
d
iag
n
o
s
is
.
W
h
ile
th
eir
f
o
c
u
s
was
o
n
u
s
ab
ilit
y
an
d
clin
ical
wo
r
k
f
lo
w
i
n
teg
r
atio
n
,
th
e
s
t
u
d
y
h
ig
h
lig
h
ts
th
e
im
p
o
r
tan
ce
o
f
tr
an
s
latin
g
AI
m
o
d
els in
to
p
r
ac
tical
to
o
ls
f
o
r
p
ath
o
lo
g
is
ts
.
I
n
s
u
m
m
ar
y
,
p
r
ev
io
u
s
s
tu
d
ies
h
ig
h
lig
h
t
th
e
p
r
o
m
is
e
o
f
d
ee
p
lear
n
in
g
an
d
en
s
em
b
le
tech
n
iq
u
es
in
ca
n
ce
r
d
ia
g
n
o
s
tics
b
u
t
also
ex
p
o
s
e
k
ey
lim
itatio
n
s
.
T
h
ese
in
clu
d
e
d
e
p
en
d
e
n
ce
o
n
o
u
td
ated
ar
ch
itectu
r
es,
p
o
o
r
g
en
er
aliza
b
ilit
y
ac
r
o
s
s
d
atasets
,
an
d
lim
ited
in
n
o
v
atio
n
in
en
s
e
m
b
le
s
tr
ateg
ies.
T
o
o
v
er
c
o
m
e
th
ese
ch
allen
g
es,
th
is
s
tu
d
y
in
tr
o
d
u
ce
s
a
n
o
v
el
en
s
em
b
le
ap
p
r
o
ac
h
b
ased
o
n
th
e
a
d
ap
tab
le
a
n
d
d
ee
p
er
VGG1
9
ar
ch
itectu
r
e,
s
p
ec
if
ically
d
esig
n
ed
f
o
r
b
r
ea
s
t c
an
ce
r
class
if
icatio
n
.
3.
M
E
T
H
O
D
T
h
e
b
r
ea
s
t
ca
n
ce
r
class
if
icatio
n
p
r
o
ce
s
s
,
illu
s
tr
ated
in
Fig
u
r
e
1
,
f
o
llo
ws
s
ev
er
al
k
e
y
s
tep
s
,
s
tar
tin
g
with
d
ata
co
llectio
n
an
d
p
r
o
g
r
ess
in
g
th
r
o
u
g
h
to
th
e
class
if
icatio
n
s
tag
e.
T
h
e
d
ata
u
s
ed
co
n
s
is
ts
o
f
im
ag
es o
f
n
o
r
m
al
an
d
ca
n
ce
r
o
u
s
b
r
ea
s
t
tis
s
u
e,
wh
ich
u
n
d
er
g
o
v
ar
io
u
s
tech
n
iq
u
es
to
en
h
a
n
ce
m
o
d
el
p
er
f
o
r
m
an
ce
,
i
n
clu
d
in
g
r
o
tatio
n
,
s
h
ea
r
,
zo
o
m
,
h
o
r
izo
n
tal
f
lip
,
f
ill
m
o
d
e
,
r
escalin
g
,
a
n
d
wid
th
a
n
d
h
ei
g
h
t
s
h
if
tin
g
.
Featu
r
e
ex
tr
ac
tio
n
is
p
er
f
o
r
m
ed
u
s
in
g
d
ee
p
lear
n
in
g
m
o
d
els
s
u
ch
as
R
esNet5
0
,
VGG1
9
,
C
o
n
v
Nex
t
B
ase,
Den
s
eNe
t1
2
1
,
E
f
f
icien
tNetV2
B
0
,
E
f
f
icien
tN
etB
0
,
Mo
b
ileNet,
an
d
NasNetM
o
b
ile.
As
s
h
o
wn
in
Fig
u
r
e
2
,
t
h
e
ar
c
h
itectu
r
e
u
tili
ze
s
th
e
VGG1
9
m
o
d
el
with
ad
d
itio
n
al
cu
s
to
m
lay
er
s
th
at
in
clu
d
e
a
f
u
lly
co
n
n
ec
ted
d
en
s
e
lay
er
with
5
1
2
u
n
its
,
f
o
llo
we
d
b
y
d
r
o
p
o
u
t
lay
e
r
s
(
with
a
d
r
o
p
o
u
t
r
ate
o
f
0
.
5
)
to
p
r
ev
en
t
o
v
er
f
itti
n
g
.
Af
ter
war
d
,
an
o
th
e
r
d
en
s
e
lay
e
r
with
2
5
6
u
n
its
an
d
a
n
o
th
er
d
r
o
p
o
u
t
lay
er
ar
e
ap
p
lied
.
T
h
e
co
n
v
o
l
u
tio
n
al
b
ase
o
f
VGG1
9
is
f
r
o
ze
n
to
p
r
eser
v
e
p
r
e
-
tr
ain
e
d
f
ea
tu
r
es,
an
d
th
e
m
o
d
if
ied
V
GG1
9
ex
tr
ac
ts
d
ee
p
f
ea
tu
r
es
f
r
o
m
t
h
e
d
ataset.
T
h
ese
ex
tr
ac
ted
f
ea
tu
r
es
ar
e
th
en
n
o
r
m
alize
d
b
ef
o
r
e
b
ein
g
f
ed
in
to
a
b
a
g
g
in
g
en
s
em
b
le
m
o
d
el.
T
h
e
b
ase
esti
m
ato
r
s
f
o
r
b
ag
g
in
g
in
cl
u
d
e
s
u
p
p
o
r
t
v
ec
to
r
class
if
ier
s
,
DT
,
an
d
LR
.
T
o
f
u
r
t
h
er
o
p
tim
ize
p
er
f
o
r
m
an
ce
,
h
y
p
er
p
ar
am
eter
tu
n
in
g
is
p
er
f
o
r
m
e
d
u
s
in
g
R
an
d
o
m
ized
Sear
ch
C
V,
en
s
u
r
in
g
th
e
b
est
co
m
b
in
atio
n
o
f
p
ar
am
eter
s
f
o
r
th
e
class
if
ier
s
.
Fig
u
r
e
1
.
Ov
e
r
v
iew
o
f
th
e
co
m
b
in
atio
n
b
etwe
en
C
NN
-
tech
n
iq
u
es a
n
d
b
ag
g
i
n
g
.
I
n
th
e
class
if
icatio
n
s
tag
e,
a
b
a
g
g
in
g
e
n
s
em
b
le
m
eth
o
d
is
em
p
lo
y
ed
with
b
ase
esti
m
ato
r
s
lik
e
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es (
SVM)
,
d
ec
is
io
n
tr
ee
s
(
DT
)
,
an
d
lo
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
to
im
p
r
o
v
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
an
d
r
o
b
u
s
tn
ess
.
T
h
is
en
s
em
b
le
ap
p
r
o
ac
h
c
o
m
b
in
es
m
u
ltip
le
p
r
e
d
ictio
n
s
f
r
o
m
d
if
f
er
en
t
b
ase
cl
ass
if
ier
s
to
g
en
er
ate
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
8
0
9
-
2
8
1
9
2812
m
o
r
e
r
eliab
le
an
d
ac
c
u
r
ate
c
lass
if
icatio
n
r
esu
lts
.
T
h
is
p
ip
elin
e
p
r
o
v
id
es
a
co
m
p
r
eh
e
n
s
iv
e
f
r
am
ewo
r
k
f
o
r
d
etec
tin
g
an
d
class
if
y
in
g
b
r
ea
s
t c
an
ce
r
im
ag
es.
T
h
is
d
iag
r
am
r
ep
r
esen
ts
a
p
r
o
p
o
s
ed
ar
c
h
itectu
r
e
f
o
r
b
r
ea
s
t
ca
n
ce
r
class
if
icatio
n
u
s
in
g
a
m
o
d
if
ied
VGG1
9
m
o
d
el
c
o
m
b
in
e
d
with
m
ac
h
in
e
lear
n
in
g
tech
n
i
q
u
es.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
th
e
d
ataset,
wh
ich
in
clu
d
es
b
r
ea
s
t
ca
n
ce
r
im
ag
es.
T
h
e
d
ata
is
s
p
lit
in
to
tr
ain
i
n
g
an
d
test
in
g
s
ets
an
d
r
esh
ap
e
d
f
o
r
co
m
p
atib
ilit
y
with
th
e
VGG1
9
m
o
d
el.
T
h
is
ar
ch
itectu
r
e
is
d
esig
n
ed
to
en
h
a
n
ce
b
o
th
ac
c
u
r
ac
y
a
n
d
r
o
b
u
s
tn
ess
in
class
if
y
in
g
b
r
e
ast
ca
n
ce
r
im
ag
es.
I
t
lev
er
ag
es
th
e
p
o
wer
f
u
l
f
ea
tu
r
e
ex
tr
ac
tio
n
ca
p
ab
ilit
ies
o
f
th
e
d
ee
p
lear
n
in
g
m
o
d
el
VGG1
9
,
co
m
b
in
ed
with
th
e
en
s
em
b
le
s
tab
ilit
y
o
f
a
b
ag
g
in
g
ap
p
r
o
ac
h
u
s
i
n
g
m
ac
h
in
e
lear
n
in
g
b
ase
esti
m
ato
r
s
.
T
h
is
in
teg
r
atio
n
aim
s
to
ca
p
italize
o
n
th
e
s
tr
en
g
th
s
o
f
b
o
th
m
eth
o
d
s
f
o
r
im
p
r
o
v
ed
d
ia
g
n
o
s
tic
p
er
f
o
r
m
a
n
ce
.
Fig
u
r
e
2
.
Ov
e
r
v
iew
o
f
th
e
m
o
d
if
ied
VGG1
9
3
.
1
.
T
he
da
t
a
s
et
des
cr
iptio
n a
nd
prepro
ce
s
s
ing
I
n
ter
n
atio
n
al
C
o
n
f
er
e
n
ce
o
n
I
m
ag
e
An
aly
s
is
an
d
R
ec
o
g
n
iti
o
n
(
I
C
I
AR
)
[
2
6
]
:
Hem
ato
x
y
lin
an
d
eo
s
in
(
H&
E
)
s
tain
ed
b
r
ea
s
t
h
is
to
lo
g
y
m
icr
o
s
co
p
y
an
d
wh
o
le
-
s
lid
e
im
ag
es
co
m
p
r
is
e
th
e
d
ataset.
T
h
e
d
atase
t
co
m
p
r
is
es 4
0
0
m
icr
o
s
co
p
e
p
ic
tu
r
es,
wh
ich
ar
e
d
is
tr
ib
u
ted
as f
o
llo
ws:
1
0
0
is
co
n
s
id
er
ed
n
o
r
m
al,
Po
s
itiv
e:
1
0
0
,
1
0
0
ca
s
es
o
f
in
s
itu
ca
r
cin
o
m
a
,
1
0
0
ca
s
es
o
f
in
v
asiv
e
ca
r
cin
o
m
a
Mic
r
o
s
co
p
y
p
ictu
r
es
ar
e
in
.
tiff
f
o
r
m
at
an
d
m
ee
t
th
e
f
o
llo
win
g
r
eq
u
ir
em
e
n
ts
:
R
ed
Gr
ee
n
B
lu
e
is
th
e
co
lo
r
m
o
d
el,
Dim
en
s
io
n
s
: 2
0
4
8
×
1
5
3
6
p
ix
els,
0
.
4
2
×
0
.
4
2
m
p
ix
el
s
ca
le,
1
0
-
2
0
MB
o
f
m
e
m
o
r
y
s
p
ac
e,
T
y
p
e
o
f
lab
el:
im
ag
e
-
wis
e.
I
n
th
is
s
tu
d
y
we
u
s
ed
th
e
two
class
es
n
o
r
m
al
an
d
b
e
n
ig
n
th
at
c
o
n
tai
n
ed
1
0
0
im
ag
es e
ac
h
.
FNAC
d
ataset:
Data
b
ase
o
f
f
i
n
e
n
ee
d
le
asp
ir
atio
n
cy
to
lo
g
y
(
FNAC
)
[
2
7
]
: We a
cq
u
ir
ed
im
ag
es o
f
th
e
FNAC
d
ataset
u
s
in
g
a
L
eica
I
C
C
5
0
HD
m
icr
o
s
co
p
e
with
4
0
0
r
eso
lu
tio
n
a
n
d
2
4
-
b
it c
o
l
o
r
d
ep
th
,
as
well
as a
5
-
m
eg
ap
ix
el
ca
m
e
r
a
attac
h
ed
t
o
t
h
e
m
icr
o
s
co
p
e
.
T
h
e
d
ig
itized
i
m
ag
es
wer
e
th
en
e
v
alu
ated
b
y
co
m
p
eten
t
ce
r
tifie
d
cy
to
p
ath
o
l
o
g
i
s
ts
,
wh
o
p
ick
e
d
2
1
2
im
a
g
es in
to
tal
1
1
3
Ma
lig
n
an
t a
n
d
9
9
B
en
ig
n
.
T
h
e
I
C
I
AR
d
ataset
was
ch
o
s
en
f
o
r
its
h
ig
h
-
q
u
ality
H&
E
-
s
tain
ed
b
r
ea
s
t
h
is
to
lo
g
y
im
ag
e
s
,
o
f
f
er
in
g
b
alan
ce
d
a
n
d
well
-
lab
eled
ca
t
eg
o
r
ies
id
ea
l
f
o
r
ea
r
ly
-
s
tag
e
c
an
ce
r
d
etec
tio
n
.
T
h
e
n
o
r
m
al
an
d
b
en
ig
n
class
es
wer
e
s
p
ec
if
ically
s
elec
ted
to
alig
n
with
th
e
s
tu
d
y
'
s
f
o
cu
s
o
n
ea
r
ly
-
s
tag
e
ab
n
o
r
m
alities
.
T
h
e
FNAC
d
ata
s
et
co
m
p
lem
en
ts
th
is
with
cy
to
lo
g
ical
-
lev
el
im
ag
in
g
,
p
r
o
v
id
in
g
f
in
e
-
g
r
ain
e
d
d
iag
n
o
s
tic
d
etails
v
alid
ated
b
y
ex
p
e
r
t
cy
to
p
ath
o
l
o
g
is
ts
.
T
o
g
eth
e
r
,
th
e
s
e
d
atasets
o
f
f
er
d
iv
e
r
s
e
an
d
r
eliab
le
d
ata
t
o
s
u
p
p
o
r
t
r
o
b
u
s
t
m
o
d
el
d
ev
elo
p
m
en
t
an
d
ev
alu
atio
n
.
3
.
2
.
Da
t
a
pre
-
pro
ce
s
s
ing
T
h
e
in
itial
an
d
m
o
s
t
im
p
o
r
tan
t
s
tep
in
b
u
ild
in
g
a
r
eliab
le
p
r
ed
ictiv
e
m
o
d
el
is
p
r
ep
r
o
ce
s
s
in
g
th
e
in
p
u
t
im
ag
es
u
s
in
g
v
ar
io
u
s
p
r
ep
r
o
c
ess
in
g
tech
n
iq
u
es.
T
o
au
g
m
en
t
an
d
b
alan
ce
th
e
d
ata,
g
eo
m
e
tr
ic
tr
an
s
f
o
r
m
atio
n
s
wer
e
ap
p
lied
.
I
n
itially
,
th
e
im
ag
es
u
n
d
er
we
n
t
H
is
to
g
r
am
E
q
u
aliza
tio
n
to
en
h
an
ce
c
o
n
tr
ast
an
d
n
o
r
m
alize
in
ten
s
ity
d
is
tr
ib
u
tio
n
.
T
h
is
w
as
f
o
llo
wed
b
y
Data
Au
g
m
en
tatio
n
to
ex
p
a
n
d
th
e
d
ataset
a
n
d
im
p
r
o
v
e
m
o
d
el
g
en
er
aliza
tio
n
.
T
h
e
au
g
m
en
tat
io
n
p
r
o
c
ess
in
clu
d
ed
tr
an
s
f
o
r
m
atio
n
s
s
u
ch
as
r
o
tatio
n
u
p
to
3
0
d
eg
r
e
es,
zo
o
m
in
g
with
a
r
an
g
e
o
f
0
.
2
,
s
h
ea
r
in
g
w
ith
a
r
a
n
g
e
o
f
0
.
2
,
wid
th
s
h
if
tin
g
with
a
r
an
g
e
o
f
0
.
3
,
h
eig
h
t
s
h
if
tin
g
with
a
r
an
g
e
o
f
0
.
3
,
h
o
r
izo
n
tal
f
lip
p
in
g
,
an
d
r
escalin
g
b
y
d
iv
id
i
n
g
p
ix
el
v
alu
es
b
y
2
5
5
.
Ad
d
itio
n
ally
,
a
f
ill
m
o
d
e
s
et
to
“
n
ea
r
est
”
was u
s
ed
to
h
an
d
le
g
ap
s
d
u
r
in
g
tr
an
s
f
o
r
m
atio
n
s
,
e
n
s
u
r
in
g
r
o
b
u
s
tn
ess
in
th
e
p
r
ep
r
o
ce
s
s
in
g
p
ip
elin
e.
T
h
e
d
ataset
was
th
en
s
p
lit
in
to
8
0
%
f
o
r
tr
ain
in
g
an
d
2
0
%
f
o
r
test
in
g
to
en
s
u
r
e
a
co
m
p
r
eh
en
s
iv
e
ev
alu
atio
n
o
f
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
T
h
is
s
p
lit
p
r
o
v
id
e
d
a
s
u
f
f
icien
t
am
o
u
n
t
o
f
d
ata
f
o
r
tr
ai
n
in
g
wh
ile
r
eser
v
in
g
an
a
p
p
r
o
p
r
iate
p
o
r
ti
o
n
f
o
r
ass
ess
in
g
th
e
m
o
d
e
l'
s
ab
ilit
y
to
g
en
er
alize
to
u
n
s
ee
n
d
ata.
A
co
n
s
is
ten
t
r
an
d
o
m
s
ee
d
was
u
s
ed
to
e
n
s
u
r
e
th
e
r
ep
r
o
d
u
cib
ilit
y
o
f
th
e
s
p
lit.
Fu
r
th
er
m
o
r
e,
class
lab
els
wer
e
p
r
ep
r
o
ce
s
s
ed
in
to
a
o
n
e
-
h
o
t
en
co
d
ed
f
o
r
m
at
to
f
ac
ilit
ate
m
u
lti
-
clas
s
cl
ass
if
icatio
n
,
en
s
u
r
in
g
co
m
p
at
ib
ili
ty
with
n
eu
r
al
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
I
mp
r
o
vin
g
b
r
ea
s
t c
a
n
ce
r
cla
s
s
ifica
tio
n
w
ith
a
n
o
ve
l V
GG1
9
-
b
a
s
ed
en
s
emb
le
…
(
C
h
a
ym
a
e
Ta
ib
)
2813
n
etwo
r
k
m
o
d
els
an
d
en
h
an
ci
n
g
th
e
ef
f
icie
n
cy
o
f
th
e
lea
r
n
in
g
p
r
o
ce
s
s
.
T
h
is
s
tan
d
ar
d
ized
p
r
ep
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
en
s
u
r
e
d
co
n
s
is
ten
cy
ac
r
o
s
s
th
e
tr
ain
in
g
a
n
d
test
in
g
p
h
ases
f
o
r
r
eliab
le
e
v
alu
atio
n
.
3
.
3
.
P
er
f
o
r
m
a
nce
m
e
a
s
ures
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
h
y
b
r
i
d
tech
n
iq
u
es
was
ass
ess
ed
u
s
i
n
g
th
e
f
o
llo
win
g
m
etr
ics
[
2
8
]
:
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all
an
d
F1
-
s
co
r
e,
d
ef
in
ed
i
n
th
e
(
1
)
-
(
4
)
.
=
+
+
+
+
(
1
)
=
+
(
2
)
=
+
(
3
)
1
=
2
×
×
+
(
4
)
wh
er
e:
T
P (
T
r
u
e
Po
s
itiv
e)
: T
h
e
n
u
m
b
e
r
o
f
ca
s
es c
o
r
r
ec
tly
p
r
ed
icted
as p
o
s
itiv
e.
T
N
(
T
r
u
e
Neg
ativ
e
)
: T
h
e
n
u
m
b
er
o
f
ca
s
es c
o
r
r
ec
tly
p
r
e
d
icted
as n
eg
ativ
e.
FP
(
Fals
e
Po
s
itiv
e)
: T
h
e
n
u
m
b
er
o
f
ca
s
es in
co
r
r
ec
tly
p
r
e
d
icted
as p
o
s
itiv
e.
FN (
Fals
e
Neg
ativ
e)
: T
h
e
n
u
m
b
er
o
f
ca
s
es in
co
r
r
ec
tly
p
r
ed
ic
ted
as n
eg
ativ
e
3
.
4
.
P
er
f
o
r
m
a
nce
s
et
up
All
ex
p
er
im
en
ts
wer
e
co
n
d
u
ct
ed
o
n
Go
o
g
le
C
o
lab
Pro
with
5
0
GB
R
AM
an
d
1
0
7
.
7
GB
d
i
s
k
s
to
r
ag
e.
T
h
e
co
d
es
wer
e
im
p
lem
en
te
d
in
T
en
s
o
r
Flo
w
u
s
in
g
Py
th
o
n
3
as
th
e
p
r
o
g
r
am
m
in
g
lan
g
u
ag
e.
T
h
e
co
m
p
u
tatio
n
al
b
ac
k
en
d
was
p
o
wer
e
d
b
y
G
o
o
g
le
C
o
m
p
u
te
E
n
g
in
e
with
C
PU
s
u
p
p
o
r
t.
T
h
e
h
y
p
er
p
a
r
a
m
eter
s
u
s
ed
in
th
e
B
ag
g
in
g
C
lass
if
ier
with
L
R
wer
e
ca
r
ef
u
lly
s
elec
ted
to
o
p
tim
ize
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
an
d
e
n
s
u
r
e
r
ep
r
o
d
u
cib
ilit
y
.
T
h
e
e
n
s
em
b
le
was
co
n
f
ig
u
r
ed
with
a
v
ar
y
in
g
n
u
m
b
er
o
f
esti
m
ato
r
s
,
s
p
ec
if
ically
5
0
,
1
0
0
,
a
n
d
2
0
0
,
to
ass
ess
t
h
e
im
p
ac
t
o
f
th
e
en
s
em
b
le
s
ize
o
n
class
if
icatio
n
p
e
r
f
o
r
m
an
ce
.
Fo
r
L
R
,
th
e
r
eg
u
lar
izatio
n
s
tr
en
g
th
(
C
)
was
tu
n
ed
o
v
er
a
r
an
g
e
o
f
v
alu
es,
in
clu
d
in
g
0
.
0
1
,
0
.
1
,
1
,
1
0
,
1
0
0
,
an
d
1
0
0
0
,
allo
win
g
f
o
r
co
n
tr
o
l
o
v
er
th
e
tr
ad
e
-
o
f
f
b
etwe
en
b
ias
an
d
v
ar
ian
ce
.
A
r
an
d
o
m
s
tate
o
f
4
2
was
u
s
ed
th
r
o
u
g
h
o
u
t
th
e
p
r
o
ce
s
s
to
g
u
ar
an
tee
co
n
s
is
ten
t
r
esu
lts
.
T
h
ese
ch
o
ices
r
ef
lect
a
s
tr
u
ctu
r
ed
an
d
m
et
h
o
d
ical
ap
p
r
o
ac
h
to
id
e
n
tify
in
g
th
e
o
p
tim
al
m
o
d
el
co
n
f
ig
u
r
atio
n
f
o
r
r
o
b
u
s
t a
n
d
r
e
liab
le
class
if
icatio
n
.
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
e
v
a
l
u
a
ti
o
n
m
e
t
r
i
c
s
a
c
c
u
r
ac
y
,
F
1
-
s
c
o
r
e
,
r
e
c
al
l
,
a
n
d
p
r
e
c
is
i
o
n
f
o
r
e
i
g
h
t
C
N
N
a
r
c
h
i
t
ec
t
u
r
es
VG
G
1
9
,
R
es
N
e
t
5
0
,
N
as
N
e
tM
o
b
i
l
e
,
M
o
b
i
l
e
N
e
t
,
E
f
f
i
ci
e
n
t
N
et
B
0
,
E
f
f
i
ci
e
n
t
N
e
tV
2
B
0
,
D
e
n
s
e
N
et
1
2
1
,
an
d
C
o
n
v
N
e
Xt
T
i
n
y
w
e
r
e
c
o
m
b
i
n
e
d
w
i
t
h
t
h
r
e
e
c
l
ass
i
f
i
e
r
s
D
T
,
L
R
,
a
n
d
S
VM
a
n
d
a
s
s
e
s
s
e
d
o
v
e
r
t
w
o
d
a
t
as
e
ts
,
FNA
C
a
n
d
I
C
I
A
R
.
T
h
e
r
e
s
u
l
ts
,
p
r
e
s
e
n
t
e
d
i
n
F
i
g
u
r
es
3
t
o
9
a
n
d
T
a
b
l
e
s
1
a
n
d
2
,
p
r
o
v
i
d
e
a
c
o
m
p
r
e
h
e
n
s
i
v
e
a
n
a
l
y
s
i
s
o
f
m
o
d
e
l
p
e
r
f
o
r
m
a
n
c
e
a
c
r
o
s
s
v
a
r
i
o
u
s
c
o
n
f
i
g
u
r
a
t
i
o
n
s
.
T
h
e
c
o
n
f
u
s
i
o
n
m
a
t
r
i
c
e
s
,
p
a
r
t
i
cu
l
a
r
l
y
f
o
r
t
h
e
F
N
A
C
d
a
t
as
e
t
as
s
h
o
w
n
i
n
F
i
g
u
r
e
s
3
,
4
,
5
,
a
n
d
9
,
d
e
m
o
n
s
t
r
a
t
e
t
h
a
t
m
o
d
e
l
s
i
n
t
e
g
r
a
t
e
d
w
it
h
SV
M
c
o
n
s
is
t
e
n
tl
y
a
c
h
i
e
v
e
t
h
e
h
i
g
h
e
s
t
a
cc
u
r
a
c
y
.
T
h
e
s
e
m
o
d
e
ls
p
r
o
d
u
c
e
s
i
g
n
i
f
i
c
a
n
t
l
y
f
ew
e
r
f
a
l
s
e
n
e
g
a
ti
v
e
s
a
n
d
f
a
ls
e
p
o
s
it
i
v
e
s
,
h
i
g
h
li
g
h
t
i
n
g
t
h
e
i
r
r
o
b
u
s
t
n
es
s
i
n
c
o
r
r
e
c
tl
y
c
l
a
s
s
i
f
y
i
n
g
c
a
n
c
e
r
o
u
s
a
n
d
n
o
n
-
c
a
n
c
e
r
o
u
s
s
a
m
p
l
es
.
F
o
r
t
h
e
I
C
I
A
R
d
a
t
as
e
t
,
F
i
g
u
r
e
6
r
e
v
e
a
l
s
a
s
i
m
il
a
r
t
r
e
n
d
,
w
h
e
r
e
C
N
Ns
c
o
m
b
i
n
e
d
wi
t
h
SV
M
o
u
t
p
e
r
f
o
r
m
D
T
a
n
d
L
R
,
d
e
m
o
n
s
t
r
a
t
i
n
g
f
e
we
r
m
i
s
c
l
as
s
i
f
i
c
at
i
o
n
s
a
n
d
s
u
p
e
r
i
o
r
c
l
a
s
s
i
f
ic
a
t
i
o
n
c
a
p
a
b
i
l
it
y
.
T
h
i
s
in
d
i
c
a
t
e
s
t
h
e
p
o
t
e
n
ti
a
l
o
f
S
V
M
t
o
e
n
h
a
n
c
e
t
h
e
d
i
a
g
n
o
s
t
i
c
a
c
c
u
r
a
c
y
o
f
C
NN
-
b
a
s
e
d
m
o
d
e
l
s
a
c
r
o
s
s
v
a
r
y
i
n
g
d
a
t
as
e
ts
.
On
th
e
FNAC
d
ataset,
T
ab
le
1
h
ig
h
lig
h
ts
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
ac
h
iev
e
d
th
e
h
ig
h
est
ac
cu
r
ac
y
o
f
9
5
%,
o
u
tp
er
f
o
r
m
in
g
NasNetM
o
b
ile
an
d
Mo
b
ileNet,
wh
ich
s
co
r
ed
ar
o
u
n
d
8
0
%,
an
d
Den
s
eNe
t1
2
1
,
wh
ich
also
s
h
o
wed
co
m
p
etitiv
e
p
er
f
o
r
m
a
n
ce
.
Fo
r
th
e
I
C
I
AR
d
ataset
s
h
o
ws
in
T
ab
le
2
,
ac
cu
r
ac
y
d
ec
l
in
ed
s
lig
h
tly
f
o
r
all
m
o
d
els,
with
VGG1
9
an
d
M
o
b
ileNet
ac
h
iev
in
g
b
etwe
en
7
0
%
an
d
7
7
%,
wh
ile
th
e
p
r
o
p
o
s
ed
m
o
d
el
s
till
o
u
tp
er
f
o
r
m
ed
o
th
er
s
with
a
n
ac
cu
r
ac
y
o
f
8
2
%.
T
h
ese
r
esu
lts
s
u
g
g
est
th
at
DT
,
th
o
u
g
h
ef
f
ec
tiv
e,
ar
e
o
u
tp
er
f
o
r
m
ed
b
y
o
th
e
r
class
if
i
er
s
in
ter
m
s
o
f
r
o
b
u
s
tn
ess
an
d
p
r
ec
is
io
n
.
L
R
ex
h
ib
ited
s
tr
o
n
g
p
er
f
o
r
m
a
n
ce
o
n
th
e
FNAC
d
ataset,
wi
th
VGG1
9
a
n
d
R
esNet5
0
ac
h
iev
in
g
ac
cu
r
ac
ies
o
f
9
2
%
an
d
8
0
%,
r
esp
ec
tiv
ely
,
as
s
h
o
wn
in
T
ab
le
1
.
Ho
wev
er
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
s
u
r
p
ass
ed
all
o
th
er
s
w
ith
a
r
em
ar
k
ab
le
ac
cu
r
ac
y
o
f
9
5
%,
in
d
icatin
g
en
h
an
ce
d
r
o
b
u
s
tn
ess
wh
en
u
s
i
n
g
th
e
in
teg
r
ated
m
eth
o
d
.
O
n
th
e
I
C
I
AR
d
ataset,
s
im
ilar
tr
e
n
d
s
wer
e
o
b
s
er
v
e
d
,
with
VGG1
9
an
d
Mo
b
ileNet
em
er
g
in
g
as
th
e
to
p
-
p
er
f
o
r
m
in
g
m
o
d
els
af
ter
th
e
p
r
o
p
o
s
ed
m
o
d
el,
wh
ich
m
ain
tain
ed
s
u
p
er
i
o
r
ac
cu
r
ac
y
an
d
r
eliab
ilit
y
.
SVM
co
n
s
is
ten
tly
d
eliv
er
ed
t
h
e
b
est
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
b
o
th
d
atasets
.
On
FNA
C
,
m
o
d
els
s
u
ch
as
VGG1
9
,
Mo
b
ileNet,
an
d
th
e
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
ac
h
iev
ed
ac
cu
r
ac
ies
r
an
g
in
g
f
r
o
m
9
0
%
t
o
9
7
%,
as
in
d
icate
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
8
0
9
-
2
8
1
9
2814
in
T
ab
le
1
.
Fo
r
I
C
I
AR
,
th
e
p
er
f
o
r
m
a
n
ce
was
s
lig
h
tly
lo
we
r
,
with
ac
cu
r
ac
ies
r
an
g
in
g
f
r
o
m
8
3
%
to
9
0
%,
as
s
h
o
wn
in
T
ab
le
2
.
T
h
e
p
r
o
p
o
s
e
d
m
o
d
el
d
em
o
n
s
tr
ated
e
x
ce
p
ti
o
n
al
ef
f
ec
tiv
e
n
ess
,
ac
h
iev
in
g
t
h
e
h
ig
h
est
ac
cu
r
ac
y
o
f
9
0
%
o
n
th
e
I
C
I
AR
d
ataset.
T
h
is
em
p
h
asizes
th
e
r
elia
b
ilit
y
o
f
th
e
c
o
m
b
in
e
d
C
NN
-
SVM
ap
p
r
o
ac
h
in
m
in
im
izin
g
f
alse
p
o
s
itiv
es
a
n
d
n
eg
ativ
es,
c
r
itical
f
o
r
ap
p
licatio
n
s
in
b
r
ea
s
t
ca
n
ce
r
d
ia
g
n
o
s
is
.
T
h
e
r
esu
lts
u
n
d
er
s
co
r
e
t
h
e
ad
v
a
n
tag
es
o
f
in
teg
r
atin
g
C
NN
ar
ch
itectu
r
es
with
SVM
c
lass
if
ier
s
,
p
ar
ticu
lar
ly
in
m
ed
ical
im
ag
e
class
if
icatio
n
task
s
wh
er
e
p
r
ec
is
io
n
an
d
r
eliab
ilit
y
a
r
e
p
ar
a
m
o
u
n
t.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
c
o
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
ed
tr
a
d
itio
n
al
s
tan
d
alo
n
e
ap
p
r
o
ac
h
es
an
d
alter
n
ativ
e
class
if
ier
co
m
b
in
atio
n
s
ac
r
o
s
s
b
o
th
d
atasets
.
T
h
ese
f
in
d
in
g
s
h
ig
h
lig
h
t
th
e
r
o
b
u
s
tn
ess
o
f
th
e
co
m
b
i
n
ed
C
NN
-
SVM
ap
p
r
o
ac
h
,
p
ar
tic
u
lar
ly
f
o
r
m
o
d
els
lik
e
VGG1
9
an
d
Mo
b
ileNet,
wh
ic
h
ex
h
i
b
it
s
u
p
er
io
r
f
ea
tu
r
e
ex
t
r
a
ctio
n
ca
p
ab
ilit
ies.
T
h
e
r
ed
u
ctio
n
in
f
alse
n
eg
ativ
es
an
d
p
o
s
itiv
es
is
esp
ec
ially
s
ig
n
if
ican
t
f
o
r
d
ia
g
n
o
s
tic
ap
p
licatio
n
s
,
wh
er
e
e
r
r
o
r
s
ca
n
h
av
e
c
r
itical
im
p
licatio
n
s
.
Ov
er
all,
th
e
in
teg
r
atio
n
o
f
C
NNs
wi
th
en
s
em
b
le
tech
n
iq
u
es
s
u
ch
as
SVM
o
f
f
er
s
a
p
r
o
m
is
in
g
p
ath
to
war
d
d
ev
elo
p
in
g
ac
cu
r
ate
an
d
r
eliab
le
d
iag
n
o
s
tic
s
y
s
tem
s
f
o
r
b
r
ea
s
t c
an
ce
r
.
Fig
u
r
e
3
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
th
e
eig
h
t CNN te
ch
n
iq
u
es c
o
m
b
in
ed
with
d
ec
is
io
n
tr
ee
Ov
e
r
FNAC
Fig
u
r
e
4
.
C
o
n
f
u
s
io
n
m
atr
ix
o
f
t
h
e
eig
h
t CNN te
ch
n
iq
u
es c
o
m
b
in
ed
with
lo
g
is
tic
r
eg
r
ess
io
n
Ov
er
FNAC
T
h
e
co
m
p
ar
is
o
n
o
f
C
NN
ar
ch
itectu
r
es
r
ev
ea
ls
v
ar
y
in
g
l
ev
els
o
f
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
if
f
er
en
t
co
n
f
ig
u
r
atio
n
s
.
VGG1
9
d
e
m
o
n
s
tr
ates
co
n
s
is
ten
tly
h
ig
h
r
esu
lts
,
ac
h
iev
in
g
a
s
tr
o
n
g
b
alan
ce
o
f
ac
cu
r
ac
y
,
F1
-
s
co
r
e,
r
ec
all,
an
d
p
r
ec
is
io
n
,
p
ar
ticu
lar
ly
wh
en
co
m
b
i
n
ed
with
SVM
an
d
b
a
g
g
in
g
,
as
h
ig
h
lig
h
ted
in
Fig
u
r
e
8
.
Mo
b
ileNet
an
d
NasNetM
o
b
ile
also
p
er
f
o
r
m
well
ac
r
o
s
s
b
o
th
d
atasets
an
d
class
if
i
er
s
,
s
h
o
wca
s
in
g
th
at
lig
h
tweig
h
t
ar
ch
itectu
r
es
ca
n
b
e
h
ig
h
ly
ef
f
ec
tiv
e
f
o
r
im
ag
e
class
if
icatio
n
task
s
.
I
n
co
n
tr
as
t,
E
f
f
icien
tNet
(
B
0
an
d
V2
B
0
)
u
n
d
e
r
p
er
f
o
r
m
s
co
m
p
ar
ed
t
o
VGG1
9
a
n
d
Mo
b
ileNet,
in
d
icatin
g
th
at
i
n
cr
ea
s
e
d
m
o
d
el
co
m
p
le
x
ity
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
I
mp
r
o
vin
g
b
r
ea
s
t c
a
n
ce
r
cla
s
s
ifica
tio
n
w
ith
a
n
o
ve
l V
GG1
9
-
b
a
s
ed
en
s
emb
le
…
(
C
h
a
ym
a
e
Ta
ib
)
2815
d
o
es
n
o
t
alwa
y
s
lead
to
b
etter
p
er
f
o
r
m
a
n
ce
in
th
is
s
p
ec
if
ic
co
n
tex
t.
Similar
ly
,
Den
s
eNe
t1
2
1
a
n
d
C
o
n
v
NeX
tTin
y
ex
h
ib
it lo
wer
p
e
r
f
o
r
m
an
ce
,
es
p
ec
ially
wh
en
p
air
e
d
with
DT
an
d
LR
.
Fig
u
r
e
5
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
eig
h
t CNN te
ch
n
iq
u
es c
o
m
b
i
n
ed
with
SVM
o
v
er
FNAC
Fig
u
r
e
6
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
eig
h
t
C
NN
tech
n
iq
u
es c
o
m
b
i
n
ed
with
DT
o
v
er
I
C
I
AR
Fig
u
r
e
7
.
C
o
n
f
u
s
io
n
m
atr
ix
o
f
t
h
e
eig
h
t CNN
tech
n
iq
u
es c
o
m
b
in
ed
with
L
R
o
v
e
r
I
C
I
AR
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
8
0
9
-
2
8
1
9
2816
Fig
u
r
e
8
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
th
e
eig
h
t CNN te
ch
n
iq
u
es c
o
m
b
in
ed
with
SVM
o
v
e
r
I
C
I
A
R
Fig
u
r
e
9
.
C
o
n
f
u
s
io
n
m
atr
i
x
o
f
th
e
m
o
d
if
ied
VGG1
9
co
m
b
in
e
d
with
th
e
b
a
g
g
in
g
o
v
er
two
d
atasets
T
ab
le
1
.
T
h
e
r
esu
lts
o
f
th
e
f
o
u
r
m
er
ics o
f
e
v
alu
atio
n
o
v
er
F
NAC
Est
i
m
a
t
o
r
C
N
N
A
C
C
F1
-
S
C
O
R
E
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
DT
V
G
G
1
9
77
%
77
%
77
%
77
%
R
e
s
n
e
t
5
0
68
%
6
7%
6
7
%
6
7
%
N
a
sN
e
t
M
o
b
i
l
e
74
%
72
%
72
%
7
6%
M
o
b
i
l
e
N
e
t
80
%
8
0
%
8
0
%
8
0
%
Ef
f
i
c
i
e
n
t
N
e
t
B
0
59
%
5
9
%
5
9
%
5
9
%
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
71
%
7
0%
7
0%
71
%
D
e
n
seN
e
t
1
2
1
77
%
7
5
%
7
5
%
7
7
%
C
o
n
v
N
e
X
t
T
i
n
y
64
%
6
2
%
6
2
%
6
3
%
P
r
o
p
o
se
d
9
5
%
9
5
%
9
5
%
9
5
%
LR
V
G
G
1
9
92
%
9
2
%
9
2
%
9
2
%
R
e
s
n
e
t
5
0
80
%
8
0
%
8
0
%
79
%
N
a
sN
e
t
M
o
b
i
l
e
91
%
9
1
%
9
1
%
9
1
%
M
o
b
i
l
e
N
e
t
96
%
9
6
%
9
6
%
9
6
%
Ef
f
i
c
i
e
n
t
N
e
t
B
0
50
%
48
%
49
%
49
%
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
93
%
9
3
%
9
3
%
9
3
%
D
e
n
seN
e
t
1
2
1
91
%
9
1
%
9
2
%
9
1
%
C
o
n
v
N
e
X
t
T
i
n
y
73
%
7
3
%
7
3
%
7
3
%
P
r
o
p
o
se
d
9
5
%
9
5
%
9
5
%
9
5
%
S
V
M
V
G
G
1
9
90
%
9
0
%
9
0
%
9
0
%
R
e
s
n
e
t
5
0
77
%
7
7
%
7
7
%
7
7
%
N
a
sN
e
t
M
o
b
i
l
e
92
%
9
2
%
9
3
%
9
2
%
M
o
b
i
l
e
N
e
t
96
%
9
6
%
9
6
%
9
6
%
Ef
f
i
c
i
e
n
t
N
e
t
B
0
55
%
5
5
%
5
5
%
5
5
%
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
73
%
7
3
%
7
3
%
7
3
%
D
e
n
seN
e
t
1
2
1
95
%
9
5
%
9
5
%
9
5
%
C
o
n
v
N
e
X
t
T
i
n
y
65
%
6
5
%
6
5
%
6
5
%
P
r
o
p
o
se
d
9
7
%
9
7
%
9
7
%
9
7
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
I
mp
r
o
vin
g
b
r
ea
s
t c
a
n
ce
r
cla
s
s
ifica
tio
n
w
ith
a
n
o
ve
l V
GG1
9
-
b
a
s
ed
en
s
emb
le
…
(
C
h
a
ym
a
e
Ta
ib
)
2817
R
eg
ar
d
in
g
class
if
ier
ef
f
ec
tiv
en
ess
,
SVM
co
n
s
is
ten
tly
em
er
g
e
s
as
th
e
b
est
-
p
er
f
o
r
m
in
g
class
if
ier
ac
r
o
s
s
b
o
th
d
atasets
,
p
ar
ticu
lar
ly
w
h
e
n
co
m
b
in
e
d
with
VGG1
9
a
n
d
Mo
b
ileNet.
I
ts
ab
ilit
y
to
ef
f
ec
t
iv
ely
h
an
d
le
h
ig
h
-
d
im
en
s
io
n
al
f
ea
t
u
r
e
s
p
ac
es
a
n
d
d
is
tin
g
u
is
h
b
etw
ee
n
class
es
lik
ely
ac
co
u
n
ts
f
o
r
its
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
.
LR
also
p
er
f
o
r
m
s
r
ea
s
o
n
a
b
ly
wel
l,
th
o
u
g
h
it
f
alls
s
h
o
r
t
o
f
SVM,
p
ar
ticu
lar
ly
wh
en
ap
p
lied
to
m
o
r
e
co
m
p
lex
ar
ch
itectu
r
es
lik
e
E
f
f
icien
tN
et.
Ho
wev
er
,
it
ac
h
iev
es
co
m
p
etitiv
e
r
esu
lts
wh
en
co
m
b
in
ed
with
VGG1
9
,
h
ig
h
lig
h
tin
g
its
p
o
ten
tial
as
a
s
im
p
ler
y
et
ef
f
ec
tiv
e
alter
n
ativ
e
in
s
o
m
e
co
n
f
ig
u
r
atio
n
s
.
T
h
ese
f
in
d
i
n
g
s
u
n
d
er
s
co
r
e
th
e
im
p
o
r
tan
ce
o
f
s
elec
tin
g
ap
p
r
o
p
r
iate
class
if
ier
-
ar
ch
itectu
r
e
co
m
b
i
n
atio
n
s
to
o
p
tim
ize
p
er
f
o
r
m
an
ce
f
o
r
s
p
ec
if
ic
d
atas
ets an
d
task
s
.
T
ab
le
2
.
T
h
e
r
esu
lts
o
f
th
e
f
o
u
r
m
etr
ics o
f
ev
al
u
atio
n
o
v
er
I
C
I
AR
Est
i
m
a
t
o
r
C
N
N
A
C
C
F1
-
S
C
O
R
E
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
DT
V
G
G
1
9
72
%
7
2
%
7
2
%
7
2
%
R
e
s
n
e
t
5
0
72
%
72
%
7
3
%
72
%
N
a
sN
e
t
M
o
b
i
l
e
72
%
7
2
%
7
2
%
7
2
%
M
o
b
i
l
e
N
e
t
70
%
7
0
%
7
0
%
7
0
%
Ef
f
i
c
i
e
n
t
N
e
t
B
0
71
%
7
0
%
7
0
%
7
3
%
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
66
%
6
6
%
6
6
%
6
6
%
D
e
n
seN
e
t
1
2
1
66
%
6
5
%
6
5
%
6
7
%
C
o
n
v
N
e
X
t
T
i
n
y
68
%
6
7
%
6
8
%
70
%
P
r
o
p
o
se
d
86
%
86
%
86
%
86
%
LR
V
G
G
1
9
83
%
8
3
%
8
3
%
8
3
%
R
e
s
n
e
t
5
0
69
%
6
9
%
6
9
%
6
9
%
N
a
sN
e
t
M
o
b
i
l
e
79
%
7
9
%
7
9
%
7
9
%
M
o
b
i
l
e
N
e
t
85
%
8
5
%
8
5
%
8
5
%
Ef
f
i
c
i
e
n
t
N
e
t
B
0
56
%
5
6
%
5
6
%
5
6
%
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
57
%
5
7
%
5
7
%
5
7
%
D
e
n
seN
e
t
1
2
1
82
%
82
%
8
3
%
82
%
C
o
n
v
N
e
X
t
T
i
n
y
77
%
7
7
%
7
7
%
7
7
%
P
r
o
p
o
se
d
90
%
9
0
%
9
0
%
9
0
%
S
V
M
V
G
G
1
9
83
%
8
3
%
8
3
%
8
3
%
R
e
s
n
e
t
5
0
68
%
6
8
%
6
8
%
6
8
%
N
a
sN
e
t
M
o
b
i
l
e
82
%
8
2
%
8
2
%
8
2
%
M
o
b
i
l
e
N
e
t
86
%
8
6
%
8
6
%
8
6
%
Ef
f
i
c
i
e
n
t
N
e
t
B
0
53
%
5
3
%
5
3
%
5
3
%
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
59
%
5
9
%
5
9
%
5
9
%
D
e
n
seN
e
t
1
2
1
86
%
8
6
%
8
6
%
8
6
%
C
o
n
v
N
e
X
t
T
i
n
y
74
%
7
4
%
7
4
%
7
4
%
P
r
o
p
o
se
d
90
%
9
0
%
9
0
%
9
0
%
5.
CO
NCLU
SI
O
N
I
n
th
is
s
tu
d
y
,
we
ap
p
lied
a
m
o
d
if
ied
VGG1
9
m
o
d
el
co
m
b
in
ed
with
b
ag
g
in
g
,
u
s
in
g
SVM
as
th
e
b
ase
class
if
ier
,
to
clas
s
if
y
im
ag
es
f
r
o
m
two
d
atasets
.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ated
s
tr
o
n
g
p
er
f
o
r
m
an
ce
,
with
h
ig
h
ac
cu
r
ac
y
an
d
co
n
s
is
ten
t
r
esu
lts
ac
r
o
s
s
tes
t
s
et
s
.
T
h
e
co
n
f
u
s
i
o
n
m
atr
ices
r
ev
ea
led
th
at
th
e
m
o
d
el
was
ef
f
ec
tiv
e
in
co
r
r
ec
tly
class
if
y
in
g
b
o
th
tr
u
e
p
o
s
itiv
es
an
d
tr
u
e
n
e
g
ativ
es,
with
m
in
im
al
m
is
class
if
icatio
n
s
.
T
h
e
h
y
p
er
p
ar
am
eter
tu
n
in
g
o
f
th
e
SVM
u
s
in
g
R
an
d
o
m
ize
d
Sear
c
h
C
V
f
u
r
th
er
o
p
tim
ized
th
e
m
o
d
el,
e
n
h
an
cin
g
its
g
en
er
aliza
tio
n
ab
ilit
y
.
T
h
ese
f
in
d
in
g
s
s
u
g
g
est
t
h
at
th
e
co
m
b
in
atio
n
o
f
d
ee
p
f
ea
tu
r
e
ex
tr
ac
tio
n
u
s
in
g
VGG1
9
an
d
th
e
en
s
em
b
le
lear
n
i
n
g
a
p
p
r
o
ac
h
p
r
o
v
id
ed
b
y
b
a
g
g
in
g
h
as
th
e
p
o
ten
tial
to
h
an
d
l
e
co
m
p
lex
im
ag
e
class
if
icatio
n
ta
s
k
s
ef
f
ec
tiv
el
y
.
Fu
tu
r
e
wo
r
k
co
u
l
d
ex
p
lo
r
e
ad
d
itio
n
al
en
h
an
ce
m
en
ts
to
im
p
r
o
v
e
m
o
d
el
p
er
f
o
r
m
an
ce
f
u
r
th
e
r
.
On
e
d
ir
ec
tio
n
co
u
l
d
in
v
o
lv
e
ex
p
e
r
im
en
tin
g
with
d
if
f
er
e
n
t
en
s
em
b
l
e
m
eth
o
d
s
,
s
u
ch
as
b
o
o
s
tin
g
,
to
co
m
p
a
r
e
its
ef
f
icac
y
with
b
ag
g
i
n
g
.
An
o
th
e
r
av
en
u
e
is
to
in
v
esti
g
ate
h
y
b
r
id
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es
b
y
in
c
o
r
p
o
r
atin
g
o
th
er
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
et
wo
r
k
s
C
NNs
alo
n
g
s
id
e
VGG1
9
,
o
r
b
y
in
teg
r
atin
g
f
ea
tu
r
e
f
u
s
io
n
tech
n
iq
u
es
t
o
c
o
m
b
in
e
i
n
f
o
r
m
atio
n
f
r
o
m
m
u
l
tip
le
lay
er
s
.
Mo
r
e
o
v
er
,
th
e
u
s
e
o
f
m
o
r
e
ad
v
a
n
ce
d
o
p
tim
izatio
n
tech
n
i
q
u
es,
s
u
ch
as B
ay
esian
o
p
tim
izatio
n
,
co
u
l
d
im
p
r
o
v
e
th
e
h
y
p
e
r
p
ar
am
eter
s
ea
r
ch
p
r
o
ce
s
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
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
8
0
9
-
2
8
1
9
2818
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
C
h
ay
m
ae
T
AI
B
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Ad
n
an
E
l A
h
m
a
d
i
✓
✓
✓
✓
Otm
an
Ab
d
o
u
n
✓
✓
✓
✓
✓
✓
✓
✓
✓
El
k
h
atir
Haim
o
u
d
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
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
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
wo
p
u
b
lic
d
atasets
wer
e
u
s
ed
in
th
is
s
tu
d
y
.
-
T
h
e
d
ata
t
h
at
s
u
p
p
o
r
t
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
a
r
e
o
p
en
l
y
av
ailab
le
in
in
Med
ica
l
I
m
a
g
e
A
n
a
lysi
s
at
h
ttp
s
://d
o
i.o
r
g
/1
0
.
1
0
1
6
/j.m
ed
i
a.
2
0
1
9
.
0
5
.
0
1
0
,
r
ef
e
r
en
ce
n
u
m
b
er
[
2
6
]
.
-
T
h
e
d
ata
th
at
s
u
p
p
o
r
t
th
e
f
i
n
d
i
n
g
s
o
f
th
is
s
tu
d
y
ar
e
o
p
en
l
y
a
v
ailab
le
in
P
a
th
o
l
o
g
y
-
R
esea
r
ch
a
n
d
P
r
a
ctice
at
h
ttp
s
://d
o
i.o
r
g
/1
0
.
1
0
1
6
/j.tic
e.
2
0
1
9
.
0
2
.
0
0
1
,
r
ef
e
r
en
ce
n
u
m
b
er
[
2
7
]
.
RE
F
E
R
E
NC
E
S
[
1
]
“
B
r
e
a
st
c
a
n
c
e
r
,
”
2
0
2
4
.
h
t
t
p
s
:
/
/
w
w
w
.
w
h
o
.
i
n
t
/
n
e
w
s
-
r
o
o
m/
f
a
c
t
-
s
h
e
e
t
s
/
d
e
t
a
i
l
/
b
r
e
a
st
-
c
a
n
c
e
r
(
a
c
c
e
ss
e
d
S
e
p
.
2
3
,
2
0
2
4
)
.
[
2
]
B
.
S
.
A
l
o
t
a
i
b
i
e
t
a
l
.
,
“
Th
e
a
c
c
u
r
a
c
y
o
f
b
r
e
a
st
c
a
n
c
e
r
d
i
a
g
n
o
s
t
i
c
t
o
o
l
s,
”
C
u
re
u
s
,
J
a
n
.
2
0
2
4
,
d
o
i
:
1
0
.
7
7
5
9
/
c
u
r
e
u
s.
5
1
7
7
6
.
[
3
]
M
.
K
a
t
a
o
k
a
,
M
.
I
i
ma,
K
.
K
.
M
i
y
a
k
e
,
a
n
d
Y
.
M
a
t
su
mo
t
o
,
“
M
u
l
t
i
p
a
r
a
me
t
r
i
c
i
m
a
g
i
n
g
o
f
b
r
e
a
st
c
a
n
c
e
r
:
A
n
u
p
d
a
t
e
o
f
c
u
r
r
e
n
t
a
p
p
l
i
c
a
t
i
o
n
s,
”
D
i
a
g
n
o
st
i
c
a
n
d
I
n
t
e
r
v
e
n
t
i
o
n
a
l
I
m
a
g
i
n
g
,
v
o
l
.
1
0
3
,
n
o
.
1
2
,
p
p
.
5
7
4
–
5
8
3
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
d
i
i
i
.
2
0
2
2
.
1
0
.
0
1
2
.
[
4
]
C
.
C
h
a
k
r
a
b
o
r
t
y
,
M
.
B
h
a
t
t
a
c
h
a
r
y
a
,
S
.
P
a
l
,
a
n
d
S
.
-
S
.
Le
e
,
“
F
r
o
m
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
t
o
d
e
e
p
l
e
a
r
n
i
n
g
:
A
d
v
a
n
c
e
s
o
f
t
h
e
r
e
c
e
n
t
d
a
t
a
-
d
r
i
v
e
n
p
a
r
a
d
i
g
m
sh
i
f
t
i
n
me
d
i
c
i
n
e
a
n
d
h
e
a
l
t
h
c
a
r
e
,
”
C
u
rre
n
t
R
e
se
a
rc
h
i
n
Bi
o
t
e
c
h
n
o
l
o
g
y
,
v
o
l
.
7
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
r
b
i
o
t
.
2
0
2
3
.
1
0
0
1
6
4
.
[
5
]
H
.
K
a
n
g
e
t
a
l
.
,
“
A
ssess
i
n
g
t
h
e
p
e
r
f
o
r
man
c
e
o
f
f
u
l
l
y
s
u
p
e
r
v
i
s
e
d
a
n
d
w
e
a
k
l
y
su
p
e
r
v
i
s
e
d
l
e
a
r
n
i
n
g
i
n
b
r
e
a
s
t
c
a
n
c
e
r
h
i
s
t
o
p
a
t
h
o
l
o
g
y
,
”
Ex
p
e
rt
S
y
st
e
m
s w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
3
7
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
3
.
1
2
1
5
7
5
.
[
6
]
C
.
Ta
i
b
,
O
.
A
b
d
o
u
n
,
a
n
d
E.
H
a
i
m
o
u
d
i
,
“
P
e
r
f
o
r
m
a
n
c
e
e
v
a
l
u
a
t
i
o
n
o
f
d
i
a
g
n
o
st
i
c
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
sy
s
t
e
m
s
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
o
n
A
p
a
c
h
e
s
p
a
r
k
,
”
i
n
L
e
c
t
u
r
e
N
o
t
e
s
i
n
Me
c
h
a
n
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
,
S
p
r
i
n
g
e
r
I
n
t
e
r
n
a
t
i
o
n
a
l
P
u
b
l
i
s
h
i
n
g
,
2
0
2
3
,
p
p
.
1
4
5
–
1
5
4
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
1
-
2
3
6
1
5
-
0
_
1
5
.
[
7
]
C
.
Ta
i
b
,
O
.
A
b
d
o
u
n
,
a
n
d
E.
K
.
H
a
i
m
o
u
d
i
,
“
O
p
t
i
mi
z
i
n
g
b
r
e
a
s
t
c
a
n
c
e
r
d
i
a
g
n
o
s
i
s:
c
o
m
b
i
n
i
n
g
h
y
b
r
i
d
a
r
c
h
i
t
e
c
t
u
r
e
s
t
h
r
o
u
g
h
A
p
a
c
h
e
S
p
a
r
k
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
E
l
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
(
I
J
EC
E)
,
v
o
l
.
1
4
,
n
o
.
4
,
p
p
.
4
2
6
1
–
4
2
7
2
,
A
u
g
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
4
i
4
.
p
p
4
2
6
1
-
4
2
7
2
.
[
8
]
C
.
Z
h
a
n
g
,
P
.
C
h
e
n
,
a
n
d
T.
Le
i
,
“
C
a
t
e
g
o
r
y
-
w
e
i
g
h
t
i
n
st
a
n
c
e
f
u
s
i
o
n
l
e
a
r
n
i
n
g
f
o
r
u
n
s
u
p
e
r
v
i
se
d
d
o
ma
i
n
a
d
a
p
t
a
t
i
o
n
o
n
b
r
e
a
st
c
a
n
c
e
r
h
i
s
t
o
p
a
t
h
o
l
o
g
y
i
ma
g
e
s,
”
Bi
o
m
e
d
i
c
a
l
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
9
9
,
Ja
n
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
sp
c
.
2
0
2
4
.
1
0
6
7
9
4
.
[
9
]
S
.
M
u
k
si
m
o
v
a
,
S
.
U
m
i
r
z
a
k
o
v
a
,
S
.
K
a
n
g
,
a
n
d
Y
.
I
.
C
h
o
,
“
C
e
r
v
i
Le
a
r
n
N
e
t
:
A
d
v
a
n
c
i
n
g
c
e
r
v
i
c
a
l
c
a
n
c
e
r
d
i
a
g
n
o
s
i
s
w
i
t
h
r
e
i
n
f
o
r
c
e
me
n
t
l
e
a
r
n
i
n
g
-
e
n
h
a
n
c
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
,
”
H
e
l
i
y
o
n
,
v
o
l
.
1
0
,
n
o
.
9
,
M
a
y
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
h
e
l
i
y
o
n
.
2
0
2
4
.
e
2
9
9
1
3
.
[
1
0
]
A
.
A
g
h
a
n
i
m,
H
.
C
h
e
k
e
n
b
a
h
,
O
.
O
u
l
h
a
j
,
a
n
d
R
.
L
a
sr
i
,
“
Q
-
Le
a
r
n
i
n
g
e
m
p
o
w
e
r
e
d
c
a
v
i
t
y
f
i
l
t
e
r
t
u
n
i
n
g
w
i
t
h
e
p
si
l
o
n
d
e
c
a
y
s
t
r
a
t
e
g
y
,
”
Pro
g
ress
I
n
El
e
c
t
r
o
m
a
g
n
e
t
i
c
s R
e
se
a
r
c
h
C
,
v
o
l
.
1
4
0
,
p
p
.
3
1
–
4
0
,
2
0
2
4
,
d
o
i
:
1
0
.
2
5
2
8
/
P
I
ER
C
2
3
1
1
1
9
0
3
.
[
1
1
]
H
.
S
l
i
m
i
,
S
.
A
b
i
d
,
a
n
d
M
.
S
a
y
a
d
i
,
“
A
d
v
a
n
c
e
d
d
e
e
p
l
e
a
r
n
i
n
g
st
r
a
t
e
g
i
e
s
f
o
r
b
r
e
a
s
t
c
a
n
c
e
r
i
m
a
g
e
a
n
a
l
y
si
s,
”
J
o
u
r
n
a
l
o
f
R
a
d
i
a
t
i
o
n
Re
se
a
rc
h
a
n
d
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
7
,
n
o
.
4
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
r
r
a
s.2
0
2
4
.
1
0
1
1
3
6
.
[
1
2
]
M
.
N
a
ss
e
r
a
n
d
U
.
K
.
Y
u
s
o
f
,
“
D
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
me
t
h
o
d
s fo
r
b
r
e
a
s
t
c
a
n
c
e
r
d
i
a
g
n
o
s
i
s:
A
s
y
st
e
ma
t
i
c
r
e
v
i
e
w
a
n
d
f
u
t
u
r
e
d
i
r
e
c
t
i
o
n
,
”
D
i
a
g
n
o
s
t
i
c
s
,
v
o
l
.
1
3
,
n
o
.
1
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
d
i
a
g
n
o
s
t
i
c
s1
3
0
1
0
1
6
1
.
[
1
3
]
M
.
A
mr
a
n
e
,
S
.
O
u
k
i
d
,
I
.
G
a
g
a
o
u
a
,
a
n
d
T.
E
n
sar
i
,
“
B
r
e
a
st
c
a
n
c
e
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
i
n
2
0
1
8
E
l
e
c
t
ri
c
El
e
c
t
ro
n
i
c
s,
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
Bi
o
m
e
d
i
c
a
l
En
g
i
n
e
e
ri
n
g
s’
M
e
e
t
i
n
g
(
EBBT)
,
A
p
r
.
2
0
1
8
,
p
p
.
1
–
4
.
d
o
i
:
1
0
.
1
1
0
9
/
E
B
B
T.
2
0
1
8
.
8
3
9
1
4
5
3
.
[
1
4
]
S
.
K
h
a
n
,
N
.
I
sl
a
m,
Z.
Ja
n
,
I
.
U
d
D
i
n
,
a
n
d
J
.
J
.
P
.
C
.
R
o
d
r
i
g
u
e
s
,
“
A
n
o
v
e
l
d
e
e
p
l
e
a
r
n
i
n
g
b
a
s
e
d
f
r
a
m
e
w
o
r
k
f
o
r
t
h
e
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
b
r
e
a
st
c
a
n
c
e
r
u
s
i
n
g
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
Pa
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
L
e
t
t
e
rs
,
v
o
l
.
1
2
5
,
p
p
.
1
–
6
,
Ju
l
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
r
e
c
.
2
0
1
9
.
0
3
.
0
2
2
.
[
1
5
]
F
.
A
t
b
a
n
,
E.
E
k
i
n
c
i
,
a
n
d
Z.
G
a
r
i
p
,
“
T
r
a
d
i
t
i
o
n
a
l
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
s
f
o
r
b
r
e
a
st
c
a
n
c
e
r
i
ma
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
w
i
t
h
o
p
t
i
mi
z
e
d
d
e
e
p
f
e
a
t
u
r
e
s
,
”
Bi
o
m
e
d
i
c
a
l
S
i
g
n
a
l
P
r
o
c
e
ssi
n
g
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
8
1
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
sp
c
.
2
0
2
2
.
1
0
4
5
3
4
.
[
1
6
]
R
.
Q
a
sr
a
w
i
e
t
a
l
.
,
“
H
y
b
r
i
d
e
n
s
e
m
b
l
e
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
f
o
r
a
d
v
a
n
c
i
n
g
b
r
e
a
s
t
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
i
n
c
l
i
n
i
c
a
l
a
p
p
l
i
c
a
t
i
o
n
s,
”
H
e
l
i
y
o
n
,
v
o
l
.
1
0
,
n
o
.
1
9
,
O
c
t
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
h
e
l
i
y
o
n
.
2
0
2
4
.
e
3
8
3
7
4
.
[
1
7
]
R
.
S
.
R
a
a
j
,
“
B
r
e
a
s
t
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
a
n
d
d
i
a
g
n
o
s
i
s
u
s
i
n
g
h
y
b
r
i
d
d
e
e
p
l
e
a
r
n
i
n
g
a
r
c
h
i
t
e
c
t
u
r
e
,
”
Bi
o
m
e
d
i
c
a
l
S
i
g
n
a
l
Pro
c
e
ssi
n
g
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
8
2
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
s
p
c
.
2
0
2
2
.
1
0
4
5
5
8
.
[
1
8
]
M
.
F
.
A
s
l
a
n
,
“
A
h
y
b
r
i
d
e
n
d
-
to
-
e
n
d
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
b
r
e
a
st
c
a
n
c
e
r
d
i
a
g
n
o
si
s
:
c
o
n
v
o
l
u
t
i
o
n
a
l
r
e
c
u
r
r
e
n
t
n
e
t
w
o
r
k
,
”
C
o
m
p
u
t
e
rs
a
n
d
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
0
5
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
e
l
e
c
e
n
g
.
2
0
2
2
.
1
0
8
5
6
2
.
[
1
9
]
V
.
S
r
e
e
p
r
a
d
a
a
n
d
D
.
K
.
V
e
d
a
v
a
t
h
i
,
“
Lu
n
g
c
a
n
c
e
r
d
e
t
e
c
t
i
o
n
f
r
o
m
X
-
r
a
y
i
m
a
g
e
s
u
si
n
g
h
y
b
r
i
d
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
,
”
Pro
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
2
3
0
,
p
p
.
4
6
7
–
4
7
4
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
3
.
1
2
.
1
0
2
.
[
2
0
]
A
.
A
.
B
a
l
a
s
u
b
r
a
ma
n
i
a
n
e
t
a
l
.
,
“
En
s
e
mb
l
e
d
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
i
ma
g
e
c
l
a
s
si
f
i
c
a
t
i
o
n
f
o
r
b
r
e
a
st
c
a
n
c
e
r
s
u
b
t
y
p
e
a
n
d
i
n
v
a
s
i
v
e
n
e
ss
d
i
a
g
n
o
si
s fr
o
m w
h
o
l
e
sl
i
d
e
i
ma
g
e
h
i
st
o
p
a
t
h
o
l
o
g
y
,
”
C
a
n
c
e
rs
,
v
o
l
.
1
6
,
n
o
.
1
2
,
Ju
n
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
c
a
n
c
e
r
s
1
6
1
2
2
2
2
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.