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
.
4
,
A
u
g
u
s
t
20
25
,
p
p
.
4
0
8
7
~
4
0
9
8
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
4
.
pp
4
0
8
7
-
4
0
9
8
4087
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Two
-
step ma
jori
t
y
vo
ting o
f
co
nv
o
lutiona
l neural n
et
wo
rks for
bra
in t
umo
r
cla
ss
ificatio
n
I
rwa
n B
u
di Sa
nto
s
o
,
Sh
o
f
f
in
Na
hwa
Ut
a
ma
,
Su
priy
o
no
I
n
f
o
r
mat
i
c
s E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
s
i
t
a
s I
sl
a
m N
e
g
e
r
i
M
a
u
l
a
n
a
M
a
l
i
k
I
b
r
a
h
i
m
,
M
a
l
a
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
1
5
,
2
0
2
4
R
ev
is
ed
Ma
r
2
5
,
2
0
2
5
Acc
ep
ted
Ma
y
2
3
,
2
0
2
5
Bra
in
t
u
m
o
r
ty
p
e
c
las
sifica
ti
o
n
is
e
ss
e
n
ti
a
l
f
o
r
d
e
term
in
in
g
f
u
rth
e
r
e
x
a
m
in
a
ti
o
n
s.
C
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
(CNN
)
m
o
d
e
l
wi
th
m
a
g
n
e
ti
c
re
so
n
a
n
c
e
ima
g
in
g
(M
RI)
i
m
a
g
e
in
p
u
t
c
a
n
imp
ro
v
e
b
r
a
in
tu
m
o
r
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
.
H
o
we
v
e
r,
d
u
e
t
o
t
h
e
h
i
g
h
ly
v
a
r
iab
le
s
h
a
p
e
,
siz
e
,
a
n
d
l
o
c
a
ti
o
n
o
f
b
ra
in
tu
m
o
rs,
in
c
re
a
sin
g
t
h
e
p
e
rf
o
rm
a
n
c
e
o
f
tu
m
o
r
c
las
sifica
ti
o
n
re
q
u
ires
c
o
n
sid
e
ra
ti
o
n
o
f
th
e
re
su
lt
s
o
f
se
v
e
ra
l
d
iffer
e
n
t
CNN
m
o
d
e
ls.
Th
e
re
fo
re
,
we
p
ro
p
o
se
d
a
two
-
ste
p
m
a
jo
rit
y
v
o
ti
n
g
(M
V)
o
n
t
h
e
re
su
lt
s
o
f
se
v
e
ra
l
CNN
m
o
d
e
ls
f
o
r
tu
m
o
r
c
las
sifica
ti
o
n
.
Th
e
CN
N
m
o
d
e
ls
in
c
lu
d
e
d
I
n
c
e
p
ti
o
n
V
3
,
Xc
e
p
t
i
o
n
,
De
n
sN
e
t2
0
1
,
Eff
icie
n
tNe
tB3
,
a
n
d
Re
sN
e
t5
0
;
e
a
c
h
wa
s
c
u
sto
m
ize
d
a
t
th
e
c
las
sifica
ti
o
n
lay
e
r.
Th
e
in
i
ti
a
l
ste
p
o
f
th
e
m
e
th
o
d
is
tran
sfe
r
-
lea
rn
i
n
g
f
o
r
e
a
c
h
CNN
m
o
d
e
l.
Th
e
n
e
x
t
ste
p
is
to
c
a
rry
o
u
t
two
ste
p
s
o
f
M
V
,
n
a
m
e
ly
M
V
o
n
t
h
e
th
re
e
CN
N
m
o
d
e
l
c
las
sifica
ti
o
n
re
su
lt
s
a
t
d
iffere
n
t
train
in
g
e
p
o
c
h
s
a
n
d
M
V
o
n
th
e
re
su
lt
s
o
f
th
e
first
ste
p
.
Th
e
p
e
rfo
rm
a
n
c
e
e
v
a
lu
a
ti
o
n
o
f
th
e
p
ro
p
o
se
d
m
e
th
o
d
u
se
d
th
e
Nic
k
p
a
rv
a
r
d
a
tas
e
t,
wh
ic
h
i
n
c
lu
d
e
d
M
RI
ima
g
e
s
o
f
g
li
o
m
a
,
p
it
u
it
a
ry
,
n
o
tu
m
o
r,
a
n
d
m
e
n
in
g
io
m
a
.
T
h
e
tes
t
re
su
lt
s
sh
o
we
d
t
h
a
t
t
h
e
p
r
o
p
o
se
d
m
e
th
o
d
o
b
tai
n
e
d
a
n
a
c
c
u
ra
c
y
o
f
9
9
.
6
9
%
with
a
p
re
c
isio
n
a
n
d
se
n
siti
v
it
y
a
v
e
ra
g
e
o
f
9
9
.
6
7
%
a
n
d
a
sp
e
c
ifi
c
it
y
o
f
9
9
.
9
0
%
.
Wi
t
h
t
h
e
se
r
e
su
lt
s,
th
e
p
r
o
p
o
s
e
d
m
e
th
o
d
is b
e
tt
e
r
th
a
n
se
v
e
ra
l
o
t
h
e
r
m
e
th
o
d
s.
K
ey
w
o
r
d
s
:
B
r
ain
tu
m
o
r
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
I
m
ag
e
Ma
g
n
etic
r
eso
n
an
ce
im
a
g
in
g
Ma
jo
r
ity
v
o
tin
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
:
I
r
wan
B
u
d
i San
to
s
o
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
,
Un
iv
er
s
itas
I
s
lam
Neg
er
i M
au
lan
a
Ma
lik
I
b
r
ah
im
J
alan
Gaja
y
an
a
5
0
Ma
lan
g
6
5
1
4
4
,
Ma
lan
g
,
I
n
d
o
n
esia
E
m
ail: ir
wan
@
ti.u
in
-
m
alan
g
.
a
c.
id
1.
I
NT
RO
D
UCT
I
O
N
T
is
s
u
e
th
at
g
r
o
ws
d
u
e
t
o
ab
n
o
r
m
al
ce
lls
in
th
e
b
r
ain
o
r
its
s
u
r
r
o
u
n
d
in
g
s
ca
n
ca
u
s
e
b
r
ain
t
u
m
o
r
s
[
1
]
.
I
n
ac
cu
r
ac
y
in
class
if
y
in
g
th
e
ty
p
e
o
f
b
r
ain
tu
m
o
r
in
a
p
er
s
o
n
ca
n
ca
u
s
e
er
r
o
r
s
in
s
u
b
s
eq
u
en
t
m
ed
ical
ac
tio
n
s
,
wh
ich
ca
n
lead
to
d
ea
th
[
2
]
,
[
3
]
.
T
o
class
if
y
th
e
ty
p
e
o
f
tu
m
o
r
,
r
ad
io
l
o
g
is
ts
u
s
u
ally
lo
o
k
at
th
e
r
esu
lts
o
f
b
r
ain
s
ca
n
s
p
r
o
d
u
ce
d
b
y
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
(
MRI)
.
MRI
im
ag
es
ca
n
m
ap
th
e
in
ter
n
al
s
tr
u
ctu
r
es
o
f
th
e
h
u
m
an
b
o
d
y
[
4
]
,
th
u
s
p
r
o
v
i
d
in
g
b
etter
v
is
u
aliza
tio
n
an
d
s
p
atial
in
f
o
r
m
atio
n
[
5
]
.
H
o
wev
er
,
m
an
u
ally
class
if
y
in
g
b
r
ain
tu
m
o
r
ty
p
es
ag
ain
s
t
MRI
im
ag
es
ca
n
ca
u
s
e
er
r
o
r
s
.
T
h
er
ef
o
r
e,
d
e
v
elo
p
in
g
an
au
to
m
atic
m
eth
o
d
is
o
n
e
s
o
lu
tio
n
to
im
p
r
o
v
e
th
e
p
er
f
o
r
m
an
ce
o
f
tu
m
o
r
ty
p
e
class
if
icatio
n
.
T
h
e
m
eth
o
d
o
f
ten
u
s
ed
to
clas
s
if
y
b
r
ain
tu
m
o
r
s
b
ased
o
n
b
r
ain
MRI
im
ag
es
is
c
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
,
as
r
e
p
o
r
te
d
in
[
6
]
–
[
8
]
.
T
h
ei
r
h
y
b
r
id
s
ch
e
m
es
p
r
o
p
o
s
ed
in
v
o
lv
ed
C
NN
an
d
c
o
n
v
e
n
tio
n
al
m
ac
h
in
e
lear
n
in
g
(
ML
)
to
o
b
tain
b
etter
b
r
ai
n
tu
m
o
r
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
h
e
u
s
e
o
f
s
ev
er
al
C
NN
m
o
d
els
in
th
e
s
ch
em
e
is
to
ex
t
r
ac
t
th
e
b
est
b
r
ain
MRI
im
ag
e
f
ea
tu
r
es
as
i
n
p
u
t
to
t
h
e
class
if
i
ca
tio
n
s
tag
e
u
s
in
g
ML
.
T
h
e
ML
is
also
u
s
ed
b
ef
o
r
e
th
e
class
if
icatio
n
s
tag
e
to
ev
alu
ate
f
ea
tu
r
es
an
d
tak
e
s
o
m
e
o
f
th
e
b
est
f
ea
tu
r
es
p
r
o
d
u
ce
d
b
y
s
ev
er
al
C
NN
m
o
d
els
[
6
]
.
T
h
e
b
est
f
e
atu
r
e
r
esu
lts
ar
e
i
n
p
u
t
i
n
to
s
e
v
er
al
ML
m
o
d
els
to
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
.
4
,
Au
g
u
s
t
20
25
:
4
0
8
7
-
4098
4088
class
if
y
b
r
ain
tu
m
o
r
s
.
T
h
e
c
o
m
b
in
atio
n
o
f
f
ea
tu
r
es
p
r
o
d
u
ce
d
b
y
ea
ch
m
o
d
el
with
s
ev
er
al
ML
s
at
th
e
class
if
icatio
n
s
tag
e
is
to
c
h
o
o
s
e
th
e
b
est
class
if
icatio
n
p
er
f
o
r
m
an
ce
r
esu
lts
am
o
n
g
th
e
co
m
b
in
atio
n
s
,
as
s
h
o
wn
in
[
7
]
.
E
n
s
u
r
i
n
g
th
at
o
n
ly
o
n
e
C
NN
m
o
d
el
is
u
s
ed
to
ex
tr
ac
t
b
r
ain
MRI
im
ag
e
f
ea
tu
r
es
an
d
co
n
tin
u
e
at
th
e
class
if
icatio
n
s
tag
e
with
o
th
e
r
m
o
d
els
also
im
p
r
o
v
es
class
if
icatio
n
p
e
r
f
o
r
m
an
ce
[
8
]
.
T
h
e
e
f
f
o
r
t
to
s
u
f
f
ice
o
n
ly
u
s
in
g
o
n
e
C
NN
m
o
d
el
is
b
ec
au
s
e
th
eir
s
tu
d
y
em
p
h
asizes
class
if
ier
o
p
tim
izatio
n
.
B
esid
es,
u
s
in
g
s
ev
er
al
C
N
N
m
o
d
els
f
o
r
MRI
im
ag
e
f
ea
tu
r
e
ex
tr
ac
tio
n
in
s
ev
e
r
al
s
tu
d
i
es
aim
s
to
o
b
t
ain
r
ep
r
esen
tat
iv
e
f
ea
tu
r
es
b
y
th
e
class
if
ier
.
Ho
wev
er
,
th
is
ef
f
o
r
t
will
o
n
ly
b
e
p
o
s
s
ib
le
if
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
a
n
d
class
if
ier
m
eth
o
d
s
ar
e
n
o
t
s
ep
ar
ated
.
T
h
en
,
th
eir
wo
r
k
w
as
to
tr
y
s
ev
er
al
m
o
d
els
f
o
r
f
e
atu
r
e
ex
t
r
ac
tio
n
an
d
co
m
b
in
e
s
ev
er
al
m
o
d
els
(
b
y
t
r
ial
an
d
er
r
o
r
)
at
th
e
class
if
icatio
n
s
tag
e
to
ch
o
o
s
e
th
e
b
est o
n
e
[
9
]
.
T
h
e
p
r
o
p
o
s
ed
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
C
NN
m
eth
o
d
s
in
s
o
m
e
p
r
ev
io
u
s
s
tu
d
ies
also
co
n
tr
ib
u
t
ed
to
th
e
p
er
f
o
r
m
an
ce
o
f
t
u
m
o
r
class
if
icatio
n
.
Pre
-
p
r
o
ce
s
s
in
g
u
s
in
g
a
Gau
s
s
ian
f
ilt
er
f
o
r
d
e
n
o
is
in
g
b
r
ain
MRI
im
a
g
es
im
p
r
o
v
e
d
C
NN
p
er
f
o
r
m
an
ce
in
tu
m
o
r
class
if
icatio
n
[
1
0
]
,
[
1
1
]
.
Ho
we
v
er
,
d
u
e
to
t
h
e
h
ig
h
ly
v
ar
iab
le
s
h
ap
e,
s
ize,
an
d
p
o
s
itio
n
o
f
tu
m
o
r
s
i
n
ea
ch
ty
p
e
o
f
b
r
ai
n
tu
m
o
r
a
n
d
th
e
co
m
p
licated
s
tr
u
ctu
r
e
o
f
th
e
h
u
m
an
b
r
ain
[
1
2
]
,
o
n
l
y
in
v
o
lv
in
g
a
C
NN
m
o
d
el
wo
u
ld
m
a
k
e
it
d
if
f
icu
l
t
to
o
b
tain
b
etter
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
T
o
o
v
er
co
m
e
th
is
p
r
o
b
le
m
,
s
ev
e
r
al
p
r
ev
io
u
s
r
esear
ch
er
s
co
m
b
in
ed
co
n
v
o
lu
tio
n
s
f
r
o
m
d
if
f
er
en
t
m
o
d
els
[
1
3
]
;
m
o
d
els
with
o
t
h
er
a
r
ch
itectu
r
es
[
2
]
,
[
1
4
]
a
n
d
s
o
m
e
co
m
b
in
ed
f
ea
tu
r
es
f
r
o
m
d
if
f
er
en
t
m
o
d
u
les
o
r
b
lo
ck
s
[
1
5
]
.
C
h
atter
jee
et
a
l.
[
1
3
]
p
r
o
p
o
s
ed
R
esNet(
1
+2
)
D,
wh
ich
co
m
b
in
es
1
D
an
d
2
D
co
n
v
o
lu
t
io
n
s
,
an
d
p
r
o
p
o
s
ed
R
esNet
m
ix
ed
co
n
v
o
lu
tio
n
,
wh
ich
co
m
b
i
n
es
2
D
an
d
3
D
co
n
v
o
lu
tio
n
s
.
T
h
eir
test
s
s
h
o
wed
th
at
p
r
o
p
o
s
ed
m
o
d
els
wer
e
s
u
p
er
io
r
to
R
esNet3
D
with
an
ac
cu
r
ac
y
o
f
9
6
.
9
8
%.
Kh
a
n
et
a
l.
[
2
]
co
m
b
i
n
ed
th
e
VGG1
6
with
th
e
2
3
-
lay
e
r
C
NN
ar
ch
itectu
r
e
to
av
o
id
o
v
er
f
itti
n
g
.
T
h
e
ev
alu
atio
n
s
h
o
wed
th
at
th
ei
r
co
m
b
in
e
d
m
o
d
el
p
r
o
d
u
ce
d
class
if
icatio
n
ac
c
u
r
ac
y
o
f
u
p
to
9
7
.
8
%
a
n
d
1
0
0
%
f
o
r
t
h
e
f
i
r
s
t
an
d
s
ec
o
n
d
d
a
ta
s
ets.
Yo
u
n
is
et
a
l.
[
1
4
]
co
n
d
u
cted
alm
o
s
t
th
e
s
a
m
e
r
esear
ch
b
y
co
m
b
in
in
g
C
NN
with
VGG1
6
a
n
d
o
b
tain
i
n
g
a
n
ac
c
u
r
ac
y
o
f
9
8
.
1
4
%.
Dif
f
er
en
t
f
r
o
m
t
h
at
u
n
d
er
tak
e
n
b
y
No
r
ee
n
et
a
l.
[
1
5
]
,
co
m
b
i
n
ed
f
ea
t
u
r
es
ex
tr
a
cted
f
r
o
m
th
e
p
r
e
-
tr
ain
ed
I
n
ce
p
tio
n
V3
m
o
d
u
le
an
d
en
ter
ed
in
to
th
e
So
f
tMa
x
.
T
h
ey
also
co
m
b
in
ed
th
e
f
ea
tu
r
es o
f
th
e
p
r
e
-
tr
ain
e
d
Den
s
Net2
0
1
b
lo
ck
s
an
d
f
o
r
w
ar
d
ed
th
em
to
So
f
tMa
x
.
T
h
e
test
r
esu
lt
s
s
h
o
wed
th
at
th
eir
p
r
o
p
o
s
ed
m
eth
o
d
s
y
ield
ed
ac
cu
r
ac
ies
o
f
9
9
.
3
4
% a
n
d
9
9
.
5
1
%,
r
esp
ec
tiv
ely
.
C
o
m
b
in
in
g
co
n
v
o
lu
tio
n
al
[
1
3
]
o
r
co
m
b
in
i
n
g
f
ea
t
u
r
es
o
r
ar
ch
itectu
r
e
[
2
]
,
[
1
4
]
,
[
1
5
]
h
as
th
e
s
am
e
p
r
in
cip
le
to
o
b
ta
in
m
o
r
e
r
e
p
r
esen
tativ
e
f
ea
tu
r
e
s
to
th
e
class
if
ier
o
r
to
p
lay
er
i
n
C
NN,
as
well
as
t
h
e
p
r
e
-
p
r
o
ce
s
s
in
g
m
et
h
o
d
r
ep
o
r
ted
in
[
1
0
]
,
[
1
1
]
,
[
1
5
]
.
Ho
w
ev
er
,
in
v
o
lv
in
g
o
n
l
y
o
n
e
o
r
two
C
NN
m
o
d
els
in
b
r
ain
tu
m
o
r
class
if
icatio
n
is
n
o
t
en
o
u
g
h
to
o
b
tain
th
e
b
est
c
l
ass
if
icatio
n
r
esu
lts
b
ec
au
s
e
b
r
ai
n
tu
m
o
r
s
h
a
v
e
h
ig
h
c
h
ar
ac
ter
is
tics
v
ar
iatio
n
in
s
h
ap
e
,
s
ize,
an
d
p
o
s
itio
n
in
MRI
im
ag
es.
C
o
m
b
in
in
g
t
h
e
class
if
icatio
n
r
esu
lts
o
f
s
ev
er
al
C
NN
m
o
d
els
th
r
o
u
g
h
a
n
en
s
em
b
le
c
o
m
b
in
atio
n
o
f
th
eir
class
if
icatio
n
r
esu
lts
ca
n
b
e
a
s
o
lu
tio
n
to
im
p
r
o
v
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
B
ased
o
n
th
e
r
esu
lts
,
p
o
ten
ti
al
p
er
f
o
r
m
an
ce
i
m
p
r
o
v
em
en
t
s
an
d
s
o
lu
tio
n
s
a
r
e
o
b
tain
ed
b
y
s
ev
er
al
p
r
ev
io
u
s
s
tu
d
ies.
C
o
m
b
in
in
g
th
e
class
if
icatio
n
r
esu
lt
s
o
f
s
ev
er
al
C
NN
m
o
d
els
i
s
o
n
e
s
o
lu
tio
n
to
im
p
r
o
v
e
tu
m
o
r
clas
s
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
On
th
e
o
th
er
h
an
d
,
c
o
n
s
id
er
in
g
t
h
e
class
if
icatio
n
r
esu
lts
o
f
m
an
y
d
if
f
er
en
t
C
NN
m
o
d
els
ca
n
o
v
er
co
m
e
h
ig
h
v
ar
iatio
n
s
in
s
h
ap
e
an
d
s
ize,
an
d
th
e
p
o
s
itio
n
o
f
th
e
tu
m
o
r
is
d
if
f
icu
lt
to
d
eter
m
in
e
[
1
6
]
.
Sev
e
r
al
well
-
k
n
o
wn
C
NN
m
o
d
els
h
a
v
e
d
if
f
er
en
t
ar
ch
itectu
r
es
with
d
if
f
er
en
t
n
u
m
b
er
s
o
f
co
n
v
o
l
u
tio
n
lay
e
r
s
an
d
co
n
v
o
lu
tio
n
f
ilter
s
izes
an
d
h
a
v
e
th
e
p
o
ten
tial
to
o
b
tain
d
if
f
er
en
t
MRI
b
r
ain
im
ag
e
f
ea
tu
r
es
an
d
class
if
icatio
n
r
esu
lts
.
T
h
u
s
,
a
class
if
icat
io
n
th
at
co
n
s
id
er
s
th
e
class
if
icat
io
n
r
esu
lts
o
f
s
ev
er
al
C
NN
m
o
d
els
in
d
ir
ec
tly
co
n
s
id
er
s
th
e
b
est
b
r
ain
MRI
im
a
g
e
f
ea
tu
r
es
o
f
ea
c
h
C
NN
m
o
d
el
s
o
th
at
it
ca
n
s
tr
en
g
th
en
th
e
f
in
al
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
[
9
]
,
[
1
7
]
,
[
1
8
]
.
I
n
a
d
d
itio
n
,
co
n
s
id
er
i
n
g
th
e
r
esu
lts
o
f
C
NN
m
o
d
els
at
s
ev
er
al
d
if
f
e
r
en
t
ep
o
ch
s
in
lear
n
in
g
is
also
ess
en
tial
to
r
ed
u
ce
u
n
ce
r
tain
ty
in
cl
ass
if
icatio
n
.
T
h
u
s
,
co
n
s
id
er
in
g
t
h
e
class
if
icatio
n
r
esu
lts
u
s
in
g
m
ajo
r
ity
v
o
tin
g
ag
ain
s
t
th
e
class
if
icatio
n
r
esu
lts
o
f
s
ev
er
al
C
NN
m
o
d
els
at
d
if
f
er
en
t
ep
o
ch
s
is
a
s
o
lu
tio
n
to
o
v
er
c
o
m
e
th
is
p
r
o
b
lem
.
T
h
e
r
ef
o
r
e,
th
is
s
tu
d
y
p
r
o
p
o
s
es
two
-
s
tep
m
ajo
r
ity
v
o
tin
g
(
MV
)
b
ased
o
n
s
ev
er
al
well
-
k
n
o
wn
C
NN
m
o
d
els
f
o
r
b
r
ai
n
tu
m
o
r
class
if
ic
atio
n
.
C
NN
m
o
d
els
in
v
o
lv
ed
in
class
if
icatio
n
in
cl
u
d
e
I
n
ce
p
tio
n
V3
[
1
9
]
,
Xce
p
ti
o
n
[
2
0
]
,
Den
s
Net2
0
1
[
2
1
]
,
E
f
f
icien
tNetB
3
[
2
2
]
,
an
d
R
esNet5
0
[
2
3
]
.
E
ac
h
C
NN
m
o
d
el
ca
n
p
r
o
v
id
e
th
e
b
est
r
esu
lts
th
r
o
u
g
h
tr
a
n
s
f
er
lear
n
in
g
,
ac
c
o
r
d
in
g
ly
,
co
n
s
id
er
in
g
th
e
class
if
icatio
n
r
esu
lts
o
f
th
ese
C
NN
m
o
d
els
th
r
o
u
g
h
m
aj
o
r
ity
v
o
tin
g
ca
n
s
tr
en
g
th
en
th
e
f
in
al
class
if
icatio
n
r
esu
lts
.
T
wo
-
s
t
ep
MV
will
s
tr
en
g
th
en
th
e
class
if
icatio
n
o
f
ea
ch
C
NN
b
y
co
n
s
id
er
i
n
g
th
e
class
if
icatio
n
r
esu
lts
at
d
if
f
er
e
n
t
ep
o
c
h
s
,
f
o
llo
we
d
b
y
th
e
cla
s
s
if
icatio
n
r
esu
lts
am
o
n
g
o
th
er
C
NN
m
o
d
els.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
s
h
o
wed
th
at
th
e
two
-
s
tep
MV
co
u
ld
im
p
r
o
v
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
b
r
ai
n
tu
m
o
r
class
if
icatio
n
.
Hen
ce
,
th
er
e
ar
e
s
ev
er
al
co
n
tr
ib
u
t
io
n
s
to
th
is
s
tu
d
y
,
wh
ich
in
clu
d
e:
a
)
in
v
o
lv
in
g
s
ev
er
al
C
NN
m
o
d
els
(
I
n
ce
p
tio
n
V
3
,
Xce
p
tio
n
,
De
n
s
Net2
0
1
,
E
f
f
icien
tNet
B
3
,
an
d
R
esNet5
0
)
b
y
a
d
ju
s
tin
g
th
e
to
p
lay
e
r
o
f
ea
ch
m
o
d
el
f
o
r
b
r
ain
tu
m
o
r
c
lass
if
icatio
n
,
an
d
b
)
p
r
o
p
o
s
in
g
a
two
-
s
tep
m
ajo
r
ity
v
o
tin
g
s
ch
em
e
to
co
m
b
in
e
th
e
class
if
icatio
n
r
esu
lts
o
f
C
NN
m
o
d
els at
d
if
f
er
e
n
t le
ar
n
in
g
ep
o
c
h
s
.
I
n
th
is
s
tu
d
y
,
Sectio
n
2
d
escr
ib
es
th
e
d
ataset
f
o
r
ev
alu
atio
n
,
th
e
s
tep
s
o
f
th
e
p
r
o
p
o
s
ed
m
e
th
o
d
,
an
d
th
e
h
y
p
er
p
a
r
am
eter
s
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
Sectio
n
3
co
n
tain
s
th
e
r
esu
lts
o
f
th
e
te
s
t
an
d
d
is
cu
s
s
es
th
e
r
esu
lts
.
Sectio
n
4
ex
p
lain
s
th
e
co
m
p
ar
is
o
n
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
s
ev
er
al
p
r
ev
io
u
s
m
eth
o
d
s
.
Fin
ally
,
s
ec
tio
n
5
co
n
tain
s
co
n
clu
s
io
n
s
b
ased
o
n
t
h
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
s
.
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
Tw
o
-
s
tep
ma
jo
r
ity
vo
tin
g
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
…
(
I
r
w
a
n
B
u
d
i S
a
n
to
s
o
)
4089
2.
M
AT
E
R
I
AL
A
ND
M
E
T
H
O
D
T
h
e
class
if
icatio
n
o
f
b
r
ain
tu
m
o
r
s
in
th
is
s
tu
d
y
is
to
class
if
y
th
e
ty
p
es
o
f
b
r
ai
n
tu
m
o
r
s
,
in
clu
d
in
g
g
lio
m
a,
m
en
in
g
io
m
a,
a
n
d
p
itu
itar
y
an
d
n
o
tu
m
o
r
.
Me
asu
r
in
g
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
i
n
th
e
class
if
icatio
n
o
f
b
r
ain
tu
m
o
r
s
is
o
n
e
o
f
t
h
e
o
b
j
ec
tiv
es
o
f
th
is
s
tu
d
y
.
Fig
u
r
e
1
s
h
o
ws
th
e
s
tep
s
o
f
b
r
ain
tu
m
o
r
class
if
icatio
n
with
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
.
T
h
e
s
tep
s
in
clu
d
e
d
ataset
ac
q
u
is
itio
n
,
p
r
e
-
p
r
o
ce
s
s
in
g
,
tr
ain
in
g
o
f
b
ase
C
NN
m
o
d
els,
an
d
two
-
s
tep
m
ajo
r
ity
v
o
ti
n
g
.
Data
s
et
ac
q
u
is
itio
n
in
v
o
l
v
es
ac
q
u
ir
in
g
MRI
im
ag
es
o
f
b
r
ain
tu
m
o
r
s
(
g
lio
m
a,
m
en
in
g
io
m
a,
a
n
d
p
itu
itar
y
)
an
d
n
o
tu
m
o
r
s
f
o
r
tr
ain
in
g
an
d
test
in
g
.
T
h
e
p
r
o
ce
s
s
also
in
v
o
lv
es
s
p
litt
in
g
th
e
t
r
ain
in
g
d
ataset
in
t
o
d
atasets
f
o
r
lear
n
in
g
an
d
m
o
d
el
v
alid
atio
n
.
T
h
e
n
ex
t
s
tep
is
p
r
e
-
p
r
o
ce
s
s
in
g
to
r
esize
ea
ch
MRI
b
r
ai
n
im
a
g
e
to
th
e
s
am
e
s
ize
f
o
r
tr
ain
in
g
an
d
test
in
g
.
All
MRI
im
ag
e
d
ata
o
f
t
h
e
s
am
e
s
ize
wer
e
f
o
r
war
d
ed
in
to
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
.
T
h
is
s
tag
e
r
eq
u
ir
ed
s
ev
er
al
h
y
p
er
p
ar
am
eter
s
in
th
e
lear
n
in
g
p
r
o
ce
s
s
o
f
s
ev
e
r
al
C
NN
m
o
d
els
to
o
b
tain
th
e
weig
h
ts
o
f
ea
ch
tr
ain
ed
m
o
d
el.
T
h
e
last
s
tep
,
b
r
ain
tu
m
o
r
class
if
icatio
n
,
u
s
es
th
e
p
r
o
p
o
s
e
d
m
eth
o
d
,
n
am
el
y
two
-
s
tep
MV
,
to
d
ev
elo
p
an
en
s
em
b
le
co
m
b
in
atio
n
o
f
th
e
f
ir
s
t p
r
ed
ictio
n
r
esu
lts
o
f
ea
c
h
C
NN
m
o
d
el
u
s
ed
.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
s
ch
e
m
e:
tu
m
o
r
class
if
icatio
n
u
s
in
g
t
wo
-
s
tep
m
ajo
r
ity
v
o
tin
g
o
f
C
NNs
2
.
1
.
Da
t
a
s
et
a
cquis
it
io
n
Data
s
et
ac
q
u
is
itio
n
in
v
o
lv
es
ac
q
u
ir
in
g
MRI
im
ag
es
o
f
b
r
ain
tu
m
o
r
s
(
g
lio
m
a
,
m
e
n
in
g
i
o
m
a,
a
n
d
p
itu
itar
y
)
an
d
n
o
tu
m
o
r
s
f
o
r
t
r
ain
in
g
an
d
test
in
g
.
T
h
e
p
r
o
c
ess
also
in
v
o
lv
es
s
p
li
ttin
g
th
e
d
ataset
f
o
r
tr
ai
n
in
g
in
to
d
atasets
f
o
r
lear
n
in
g
an
d
v
alid
atio
n
.
I
n
t
h
is
s
tu
d
y
,
b
r
ain
MRI
im
ag
es
(
ax
ial,
s
ag
ittal,
a
n
d
co
r
o
n
al
p
lan
es)
f
o
r
test
in
g
wer
e
o
b
tain
ed
f
r
o
m
th
e
Nick
p
ar
v
ar
b
r
ain
tu
m
o
r
MRI
d
ataset
[
2
4
]
.
T
h
e
d
atase
t
was
o
b
tain
ed
f
r
o
m
th
e
Kag
g
le
web
s
ite,
wh
ich
co
n
tain
s
a
co
m
b
in
atio
n
o
f
th
e
Sa
r
taj
[
2
5
]
,
B
R
3
5
H
[
2
6
]
,
a
n
d
Fig
s
h
ar
e
[
2
7
]
d
atasets
.
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
.
4
,
Au
g
u
s
t
20
25
:
4
0
8
7
-
4098
4090
T
h
e
d
ataset
co
n
s
is
ts
o
f
d
atas
ets
f
o
r
tr
ain
in
g
an
d
test
in
g
t
o
ev
alu
ate
th
e
class
if
icatio
n
m
eth
o
d
.
T
h
e
to
tal
n
u
m
b
er
o
f
b
r
ain
MRI
im
ag
e
s
in
th
e
tr
ain
in
g
d
ataset
was
5
,
7
1
2
,
in
clu
d
i
n
g
1
,
3
2
1
g
lio
m
a
im
ag
es,
1
,
3
3
9
m
en
in
g
io
m
a
im
a
g
es,
1
,
4
5
7
p
i
tu
itar
y
im
ag
es,
an
d
1
,
5
9
5
n
o
-
tu
m
o
r
im
ag
es.
Me
an
wh
ile,
th
e
to
tal
n
u
m
b
er
o
f
b
r
ain
MRI
im
ag
es
f
o
r
test
in
g
is
1
,
3
1
1
,
in
clu
d
in
g
3
0
0
g
lio
m
a
im
ag
es,
3
0
6
m
en
in
g
io
m
a
im
ag
es,
3
0
0
p
itu
itar
y
im
ag
es,
an
d
4
0
5
n
o
-
tu
m
o
r
im
ag
es.
Fig
u
r
e
2
s
h
o
ws
g
lio
m
a
,
m
en
in
g
io
m
a
,
p
itu
itar
y
,
an
d
n
o
-
tu
m
o
r
b
r
ai
n
MRI
im
ag
es.
T
o
e
v
alu
ate
ea
c
h
b
as
e
C
NN
m
o
d
el,
1
0
%
(
5
7
2
b
r
a
in
MRI
im
ag
es)
wer
e
r
an
d
o
m
ly
tak
en
f
r
o
m
th
e
o
r
ig
in
al
tr
a
in
in
g
d
ataset
f
o
r
v
alid
atio
n
.
T
h
u
s
,
th
er
e
wer
e
5
7
2
MRI
im
ag
es
f
o
r
v
alid
atio
n
an
d
th
e
r
em
ai
n
in
g
5
,
1
4
0
MRI
im
ag
es
f
o
r
th
e
lear
n
in
g
p
r
o
ce
s
s
o
f
ea
c
h
b
ase
C
NN
m
o
d
el.
T
ab
le
1
s
h
o
ws
th
e
co
m
p
o
s
itio
n
o
f
th
e
d
ataset
u
s
ed
f
o
r
th
e
tr
ain
in
g
,
test
in
g
,
an
d
v
alid
atio
n
p
r
o
ce
s
s
es.
Fig
u
r
e
2
.
E
x
am
p
les o
f
g
lio
m
a,
m
en
in
g
io
m
a,
p
itu
itar
y
,
an
d
n
o
tu
m
o
r
in
MRI
im
ag
es
T
ab
le
1
.
T
h
e
co
m
p
o
s
itio
n
o
f
tu
m
o
r
(
g
lio
m
a,
m
en
in
g
i
o
m
a,
p
i
tu
itar
y
)
an
d
n
o
tu
m
o
r
f
o
r
ev
al
u
atio
n
o
n
th
e
Nick
p
ar
v
ar
d
ataset
La
b
e
l
Tr
a
i
n
i
n
g
V
a
l
i
d
a
t
i
o
n
Te
st
i
n
g
M
R
I
P
l
a
n
e
g
l
i
o
m
a
1
,
1
8
9
1
3
2
3
0
0
A
x
i
a
l
,
S
a
g
i
t
t
a
l
,
a
n
d
C
o
r
o
n
a
l
men
i
n
g
i
o
m
a
1
,
2
0
5
1
3
4
3
0
6
A
x
i
a
l
,
S
a
g
i
t
t
a
l
,
a
n
d
C
o
r
o
n
a
l
p
i
t
u
i
t
a
r
y
1
,
3
1
1
1
4
6
3
0
0
A
x
i
a
l
,
S
a
g
i
t
t
a
l
,
a
n
d
C
o
r
o
n
a
l
n
o
t
u
m
o
r
1
,
4
3
5
1
6
0
4
0
5
A
x
i
a
l
,
S
a
g
i
t
t
a
l
,
a
n
d
C
o
r
o
n
a
l
To
t
a
l
5
,
1
4
0
5
7
2
1
3
1
1
2
.
2
.
P
re
-
pro
ce
s
s
ing
o
f
bra
in M
RI im
a
g
e
I
n
th
is
s
tu
d
y
,
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
is
u
s
ed
to
r
esize
ea
c
h
MRI
b
r
ain
im
a
g
e
to
th
e
s
am
e
s
ize
f
o
r
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
in
g
.
I
n
th
e
p
r
o
p
o
s
ed
s
ch
em
e
in
Fig
u
r
e
1
,
all
MRI
im
ag
es
in
v
o
lv
ed
as
in
p
u
t
to
th
e
b
as
e
C
NN
m
o
d
els
ar
e
s
ized
2
2
4
×
2
2
4
×
3
.
E
ac
h
MRI
b
r
ain
im
ag
e
is
an
R
GB
im
ag
e
with
th
r
ee
ch
an
n
els,
n
am
ely
r
e
d
,
g
r
ee
n
,
an
d
b
lu
e,
with
ea
ch
ch
a
n
n
el
m
ea
s
u
r
in
g
2
2
4
×
2
2
4
p
ix
els.
T
h
e
in
p
u
t
s
h
a
p
e
o
f
th
e
MRI
b
r
ain
im
ag
e
is
th
e
r
ec
o
m
m
e
n
d
atio
n
o
f
ea
ch
b
ase
C
NN
m
o
d
el,
as
s
h
o
wn
in
T
ab
le
2
.
Fu
r
th
er
m
o
r
e,
all
r
esized
b
r
ain
MRI
im
ag
es
in
th
e
tr
ain
in
g
a
n
d
v
alid
atio
n
d
atasets
ca
n
b
e
f
o
r
war
d
ed
to
t
h
e
lear
n
in
g
p
r
o
ce
s
s
.
T
h
e
s
am
e
th
in
g
ap
p
lies
to
th
e
test
in
g
d
ataset;
af
ter
b
ein
g
r
esized
,
ea
ch
M
R
I
b
r
ain
test
in
g
im
ag
e
ca
n
b
e
en
ter
ed
in
to
th
e
class
if
icatio
n
p
r
o
ce
s
s
.
2
.
3
.
T
ra
ini
ng
o
f
b
a
s
e
CNN
m
o
dels
T
h
is
s
ec
tio
n
co
v
er
s
th
e
lear
n
in
g
an
d
h
y
p
er
p
ar
am
eter
s
o
f
ea
ch
b
ase
C
NN
m
o
d
el
f
o
r
en
s
em
b
le
co
m
b
in
atio
n
in
th
e
n
ex
t
s
tag
e.
T
h
e
b
ase
C
NN
m
o
d
els
in
th
is
s
tu
d
y
in
clu
d
e
Xce
p
tio
n
,
I
n
ce
p
tio
n
V3
,
Den
s
Net2
0
1
,
E
f
f
icien
tNetB
3
,
an
d
R
esNet5
0
.
B
ef
o
r
e
lear
n
in
g
,
ea
ch
C
NN
m
o
d
el
o
n
th
e
to
p
lay
er
n
ee
d
s
cu
s
to
m
izatio
n
to
class
if
y
f
o
u
r
class
es
:
g
lio
m
a,
m
en
in
g
io
m
a,
p
it
u
itar
y
,
an
d
n
o
tu
m
o
r
.
T
o
o
b
tain
g
o
o
d
lear
n
i
n
g
r
esu
lts
,
th
e
lear
n
in
g
p
r
o
ce
s
s
o
f
ea
ch
b
ase
m
o
d
el
r
e
q
u
ir
es
h
y
p
er
p
a
r
am
eter
tu
n
in
g
.
T
ab
le
2
p
r
o
v
id
es
an
o
v
er
v
iew
o
f
th
e
h
y
p
er
p
ar
a
m
et
er
s
ettin
g
s
f
o
r
ea
c
h
b
ase
m
o
d
e
l,
in
clu
d
i
n
g
in
p
u
t
s
h
ap
e,
cu
s
to
m
izatio
n
o
n
t
h
e
to
p
lay
er
,
ep
o
ch
,
b
atc
h
s
ize,
lo
s
s
f
u
n
ctio
n
,
o
p
tim
izer
,
lear
n
in
g
r
ate,
an
d
m
o
d
el
weig
h
t
in
itializatio
n
.
I
n
tr
ain
in
g
an
d
test
in
g
,
th
e
in
p
u
t
s
h
ap
e
o
f
ea
ch
b
ase
m
o
d
el
is
22
4×
2
2
4
×
3
.
T
h
e
in
p
u
t
s
h
ap
e
is
th
e
r
ec
o
m
m
en
d
atio
n
o
f
th
e
m
o
d
el
in
g
en
er
al.
Ad
j
u
s
tm
en
ts
to
th
e
to
p
lay
e
r
f
o
r
tu
m
o
r
c
lass
if
icatio
n
in
v
o
lv
e
all
b
ase
C
NN
m
o
d
els.
T
h
e
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
Tw
o
-
s
tep
ma
jo
r
ity
vo
tin
g
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
…
(
I
r
w
a
n
B
u
d
i S
a
n
to
s
o
)
4091
I
n
ce
p
tio
n
V3
,
Den
s
Net2
0
1
,
a
n
d
E
f
f
icien
tNetB
3
m
o
d
els
u
s
e
g
lo
b
al
av
er
a
g
e
p
o
o
lin
g
an
d
d
en
s
e
s
ize
4
(
m
an
y
class
es
/lab
els)
with
a
So
f
tMa
x
ac
tiv
atio
n
f
u
n
ctio
n
o
n
th
e
to
p
lay
er
.
Me
an
wh
ile,
a
d
ju
s
ti
n
g
t
h
e
Xce
p
tio
n
m
o
d
el
o
n
th
e
to
p
lay
er
also
in
v
o
l
v
es
g
lo
b
al
av
e
r
ag
e
p
o
o
lin
g
,
b
u
t
th
e
o
u
tp
u
t
lay
er
u
s
es
a
s
ig
m
o
id
ac
tiv
atio
n
f
u
n
ctio
n
[
2
8
]
.
T
h
e
u
s
e
o
f
th
e
m
o
d
el'
s
g
lo
b
al
av
er
a
g
e
p
o
o
lin
g
an
d
ac
tiv
atio
n
f
u
n
ctio
n
h
as
b
ee
n
p
r
o
v
en
to
b
e
r
o
b
u
s
t
in
p
r
ac
tice
an
d
is
r
ec
o
m
m
e
n
d
ed
b
y
th
e
m
o
d
el
[
2
9
]
–
[
3
1
]
.
I
n
R
esNet5
0
,
th
e
to
p
lay
er
o
f
t
h
e
m
o
d
el
in
v
o
lv
es
av
er
ag
e
p
o
o
lin
g
a
n
d
f
latten
l
ay
er
s
with
a
So
f
tMa
x
ac
tiv
at
io
n
f
u
n
ctio
n
o
n
th
e
o
u
tp
u
t
la
y
er
.
T
h
e
u
s
e
o
f
th
e
ac
tiv
atio
n
f
u
n
cti
o
n
is
also
r
ec
o
m
m
en
d
ed
b
y
t
h
e
C
NN
m
o
d
el
in
[
9
]
a
n
d
[
3
2
]
.
T
ab
le
2
.
Hy
p
er
p
ar
a
m
eter
tu
n
i
n
g
o
f
b
ase
C
NN
m
o
d
els
H
y
p
e
r
p
a
r
a
me
t
e
r
B
a
se
C
N
N
m
o
d
e
l
s
I
n
c
e
p
t
i
o
n
V
3
X
c
e
p
t
i
o
n
D
e
n
sN
e
t
2
0
1
Ef
f
i
c
i
e
n
t
N
e
t
B
3
R
e
sN
e
t
5
0
I
n
p
u
t
sh
a
p
e
2
2
4
x
2
2
4
x
3
2
2
4
x
2
2
4
x
3
2
2
4
x
2
2
4
x
3
2
2
4
x
2
2
4
x
3
2
2
4
x
2
2
4
x
3
To
p
l
a
y
e
r
G
l
o
b
a
l
A
v
e
r
a
g
e
P
o
o
l
i
n
g
D
e
n
se
(
4
,
a
c
t
i
v
a
t
i
o
n
=
'S
o
f
t
M
a
x
')
G
l
o
b
a
l
A
v
e
r
a
g
e
P
o
o
l
i
n
g
D
e
n
se
(
4
,
a
c
t
i
v
a
t
i
o
n
=
'si
g
m
o
i
d
')
G
l
o
b
a
l
A
v
e
r
a
g
e
P
o
o
l
i
n
g
D
e
n
se
(
4
,
a
c
t
i
v
a
t
i
o
n
=
'S
o
f
t
M
a
x
')
G
l
o
b
a
l
A
v
e
r
a
g
e
P
o
o
l
i
n
g
D
e
n
se
(
4
,
a
c
t
i
v
a
t
i
o
n
=
'S
o
f
t
M
a
x
')
A
v
e
r
a
g
e
P
o
o
l
i
n
g
F
l
a
t
t
e
n
D
e
n
se
(
4
,
a
c
t
i
v
a
t
i
o
n
=
'S
o
f
t
M
a
x
'
)
Ep
o
c
h
1
0
,
2
0
,
3
0
1
0
,
2
0
,
3
0
1
0
,
2
0
,
3
0
1
0
,
2
0
,
3
0
1
0
,
2
0
,
3
0
B
a
t
c
h
si
z
e
16
16
16
16
16
Le
a
r
n
i
n
g
r
a
t
e
0
.
0
0
1
0
.
0
0
1
0
.
0
0
1
0
.
0
0
1
0
.
0
0
1
Lo
ss f
u
n
c
t
i
o
n
C
a
t
e
g
o
r
i
c
a
l
C
r
o
ss
e
n
t
r
o
p
y
C
a
t
e
g
o
r
i
c
a
l
C
r
o
ss
e
n
t
r
o
p
y
C
a
t
e
g
o
r
i
c
a
l
C
r
o
ss
e
n
t
r
o
p
y
C
a
t
e
g
o
r
i
c
a
l
C
r
o
ss
e
n
t
r
o
p
y
C
a
t
e
g
o
r
i
c
a
l
C
r
o
ss
e
n
t
r
o
p
y
O
p
t
i
mi
z
e
r
A
d
a
ma
x
A
d
a
ma
x
A
d
a
ma
x
A
d
a
ma
x
A
d
a
ma
x
W
e
i
g
h
t
i
n
i
t
i
a
l
i
z
a
t
i
o
n
I
mag
e
N
e
t
I
mag
e
N
e
t
I
mag
e
N
e
t
I
mag
e
N
e
t
I
mag
e
N
e
t
E
ac
h
b
ase
C
NN
m
o
d
el
in
th
is
s
tu
d
y
in
v
o
lv
ed
lear
n
in
g
at
ep
o
ch
s
1
0
,
2
0
,
a
n
d
3
0
.
L
ea
r
n
i
n
g
a
t
d
if
f
er
e
n
t
ep
o
ch
s
ac
co
m
m
o
d
ates
th
e
p
o
s
s
ib
ilit
y
o
f
v
ar
y
in
g
class
if
icatio
n
r
esu
lts
in
ea
ch
m
o
d
el.
E
ac
h
b
ase
C
NN
m
o
d
el's
b
atch
s
ize
an
d
lear
n
i
n
g
r
ate
a
r
e
1
6
an
d
0
.
0
0
1
.
T
h
e
lear
n
in
g
in
v
o
lv
i
n
g
b
atch
s
ize
a
n
d
lear
n
in
g
r
ate
s
izes
h
as
s
h
o
wn
th
e
b
est
p
er
f
o
r
m
an
ce
r
esu
lts
[
1
7
]
,
[
1
8
]
,
[
3
3
]
.
T
h
e
o
p
tim
izer
en
g
ag
ed
in
th
e
lear
n
in
g
p
r
o
ce
s
s
is
Ad
am
ax
,
wh
ich
s
h
o
ws m
o
r
e
s
tab
le
r
esu
lts
[
3
4
]
,
[
3
5
]
.
T
h
is
s
tu
d
y
'
s
class
if
icatio
n
o
f
b
r
ain
t
u
m
o
r
s
is
m
u
lti
-
class
,
h
en
ce
,
th
e
l
o
s
s
f
u
n
ctio
n
u
s
ed
in
th
e
lear
n
in
g
is
ca
teg
o
r
ical
cr
o
s
s
-
en
tr
o
p
y
[
3
0
]
,
[
3
6
]
.
I
n
a
d
d
itio
n
,
th
e
lear
n
in
g
o
f
ea
ch
C
NN
m
o
d
el
ap
p
lies
tr
an
s
f
er
lear
n
in
g
.
I
n
th
is
s
tu
d
y
,
th
e
in
itial
weig
h
ts
wer
e
tak
en
f
r
o
m
th
e
r
esu
lts
o
f
lear
n
in
g
b
ase
C
NN
m
o
d
els
u
s
in
g
th
e
I
m
ag
eNe
t
d
ataset
[
3
7
]
,
th
u
s
h
el
p
in
g
th
e
lear
n
in
g
p
r
o
ce
s
s
o
f
C
NN
m
o
d
els o
n
s
m
all
d
atasets
[
3
8
]
.
I
n
ce
p
tio
n
V3
is
a
C
NN
ar
ch
ite
ctu
r
e
f
r
o
m
th
e
I
n
ce
p
tio
n
f
am
i
ly
with
s
ev
er
al
im
p
r
o
v
em
en
ts
[
1
9
]
.
T
h
e
m
o
d
el
ar
ch
itectu
r
e
h
as
a
s
team
b
lo
ck
an
d
in
ce
p
tio
n
b
lo
c
k
(
A,
B
,
an
d
C
)
,
r
ed
u
ctio
n
b
lo
ck
(
A
an
d
B
)
,
a
n
d
au
x
iliar
y
class
if
ier
b
lo
ck
.
I
n
th
is
s
tu
d
y
,
th
e
I
n
ce
p
tio
n
V
3
lear
n
in
g
p
r
o
ce
s
s
in
v
o
lv
ed
t
h
r
e
e
lear
n
i
n
g
ep
o
ch
s
,
n
am
ely
1
0
,
2
0
,
a
n
d
3
0
.
As
a
r
esu
lt,
th
er
e
ar
e
3
I
n
ce
p
tio
n
V
3
class
if
icatio
n
r
esu
lts
,
n
am
el
y
at
ep
o
ch
1
0
(
ℎ
1
)
,
ep
o
ch
2
0
(
ℎ
2
)
,
an
d
ep
o
c
h
3
0
(
ℎ
3
)
in
Fig
u
r
e
1
.
Xce
p
tio
n
is
a
C
NN
ar
ch
itect
u
r
e
th
at
co
n
tin
u
es
th
e
I
n
ce
p
t
io
n
ar
c
h
itectu
r
e
[
2
0
]
.
Xce
p
tio
n
h
as
3
6
co
n
v
o
l
u
tio
n
al
lay
er
s
d
iv
i
d
ed
i
n
to
1
4
m
o
d
u
les.
T
h
e
ar
ch
itect
u
r
e
h
as
a
co
n
v
o
lu
tio
n
o
p
er
ati
o
n
(
f
ilter
s
ize
1
×
1
)
to
o
b
tain
cr
o
s
s
-
ch
an
n
el
co
r
r
el
atio
n
as
a
2
D
+
1
D
m
a
p
p
in
g
.
Xce
p
tio
n
in
th
is
s
tu
d
y
also
in
v
o
lv
ed
t
h
r
ee
lear
n
in
g
ep
o
ch
s
,
n
am
el
y
1
0
,
2
0
,
a
n
d
3
0
.
T
h
e
class
if
icatio
n
r
esu
lts
u
s
in
g
Xce
p
tio
n
at
ep
o
c
h
ten
a
r
e
s
h
o
wn
with
ℎ
4
,
ep
o
ch
2
0
with
ℎ
5
,
an
d
e
p
o
ch
3
0
with
ℎ
6
in
Fig
u
r
e
1
.
Den
s
Net2
0
1
is
a
C
NN
ar
ch
ite
ctu
r
e
with
ea
ch
lay
er
to
ea
ch
n
ex
t
lay
er
co
n
n
ec
te
d
u
s
in
g
f
e
ed
f
o
r
war
d
[
2
1
]
.
T
h
u
s
,
t
h
e
ar
c
h
itectu
r
e
h
as
d
ir
ec
t
co
n
n
ec
tio
n
s
o
f
(
L
×(
L
+1
)
)
/2
,
with
ea
ch
la
y
er
p
r
o
d
u
cin
g
a
f
ea
tu
r
e
m
a
p
as
in
p
u
t
to
th
e
n
ex
t
la
y
er
.
T
h
e
Den
s
eNe
t
-
2
0
1
ar
ch
itectu
r
e
h
as
f
o
u
r
d
en
s
e
b
lo
ck
s
,
ea
c
h
wit
h
6
,
1
2
,
2
4
,
a
n
d
1
6
co
n
v
o
l
u
tio
n
b
lo
c
k
s
.
I
n
t
h
is
s
t
u
d
y
,
Den
s
eNe
t
-
2
0
1
also
ap
p
lied
th
r
ee
lear
n
i
n
g
ep
o
ch
s
,
n
am
ely
1
0
,
2
0
,
a
n
d
3
0
,
with
th
e
class
if
icatio
n
r
esu
lts
o
f
ea
ch
e
p
o
ch
i
n
d
icate
d
with
ℎ
7
,
ℎ
8
,
an
d
ℎ
9
(
s
ee
Fig
u
r
e
1
)
.
E
f
f
icien
tNetB
3
is
o
n
e
o
f
th
e
C
NN
ar
ch
itectu
r
es
p
ar
t
o
f
t
h
e
E
f
f
icien
tNet
v
ar
ia
n
t
th
at
s
ca
les
al
l
d
im
en
s
io
n
s
u
n
if
o
r
m
ly
u
s
in
g
p
o
o
lin
g
co
e
f
f
icien
ts
[
2
2
]
.
T
h
e
C
NN
ar
ch
itectu
r
e
h
as
two
co
n
v
o
lu
tio
n
lay
e
r
s
an
d
s
ev
en
m
o
b
ile
b
o
ttlen
ec
k
c
o
n
v
o
lu
tio
n
(
MBC
o
n
v
)
la
y
er
b
l
o
c
k
s
.
E
ac
h
lay
er
in
th
e
MBC
o
n
v
b
lo
ck
,
e
x
ce
p
t
t
h
e
f
ir
s
t
lay
er
,
in
v
o
lv
es
an
in
v
er
t
ed
r
esid
u
al
co
n
n
ec
tio
n
.
E
f
f
ici
en
tNetB
3
also
ap
p
lied
th
r
ee
ep
o
ch
s
in
lear
n
in
g
with
th
e
class
if
icatio
n
r
esu
lts
s
h
o
wn
b
y
ℎ
10
at
ep
o
ch
1
0
,
ℎ
11
at
ep
o
c
h
2
0
,
a
n
d
ℎ
12
at
ep
o
c
h
3
0
in
Fig
u
r
e
1
.
R
esNet5
0
is
a
C
NN
ar
ch
itectu
r
e
th
at
u
s
es
r
esid
u
al
c
o
n
n
ec
tio
n
s
s
o
th
at
th
e
n
etwo
r
k
ca
n
lea
r
n
a
s
er
ies
o
f
r
esid
u
al
f
u
n
ctio
n
s
th
at
m
ap
in
p
u
ts
to
th
e
d
esire
d
o
u
tp
u
ts
[
2
3
]
.
T
h
e
ar
ch
itectu
r
e
h
as
a
co
n
v
o
lu
ti
o
n
al
lay
er
,
an
id
en
tity
b
lo
ck
,
an
d
a
co
n
v
o
lu
tio
n
al
b
lo
c
k
.
T
h
e
co
n
v
o
l
u
tio
n
al
lay
er
is
im
p
lem
en
ted
to
ex
tr
ac
t
f
ea
tu
r
es,
wh
ile
th
e
id
en
tity
an
d
co
n
v
o
lu
tio
n
al
b
lo
ck
s
tr
an
s
f
o
r
m
f
ea
tu
r
es.
R
esNet5
0
al
s
o
in
v
o
lv
ed
th
r
ee
ep
o
ch
s
in
lear
n
in
g
f
r
o
m
wh
ich
class
if
icatio
n
r
esu
lts
ar
e
s
h
o
wn
as
ℎ
13
in
ep
o
ch
1
0
,
ℎ
14
in
ep
o
ch
2
0
,
an
d
ℎ
15
in
ep
o
ch
3
0
in
Fig
u
r
e
1
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
.
4
,
Au
g
u
s
t
20
25
:
4
0
8
7
-
4098
4092
2
.
4
.
T
wo
-
s
t
ep
m
a
j
o
rit
y
v
o
t
ing
Ma
jo
r
ity
v
o
tin
g
(
MV
)
in
im
p
lem
en
tatio
n
h
as
an
im
p
ac
t
o
n
im
p
r
o
v
in
g
class
if
icatio
n
o
r
d
etec
tio
n
p
er
f
o
r
m
an
ce
[
9
]
,
[
1
7
]
,
[
1
8
]
.
T
wo
-
s
tep
m
ajo
r
ity
v
o
tin
g
is
a
p
r
o
p
o
s
ed
m
eth
o
d
t
h
at
in
v
o
lv
es
two
s
tep
s
o
f
th
e
m
ajo
r
ity
v
o
tin
g
p
r
o
ce
s
s
.
T
h
e
f
ir
s
t
s
tep
,
MV
(
o
n
e
-
s
tep
MV
)
,
is
p
er
f
o
r
m
ed
o
n
th
e
class
if
icatio
n
r
esu
lts
o
f
th
e
s
am
e
C
NN
m
o
d
el
with
d
if
f
e
r
e
n
t
lear
n
in
g
e
p
o
ch
s
.
(
1
)
,
(
2
)
,
(
3
)
,
(
4
)
,
an
d
(
5
)
ar
e
m
ath
e
m
atica
l
eq
u
atio
n
s
o
f
th
e
f
ir
s
t
s
tep
MV
f
o
r
th
e
I
n
ce
p
tio
n
V3
,
Xce
p
tio
n
,
Den
s
Net2
0
1
,
E
f
f
icien
tNe
tB
3
,
an
d
R
esNet5
0
m
o
d
els.
T
h
e
f
ir
s
t
s
tep
,
MV
,
aim
s
to
s
tr
en
g
th
en
th
e
class
if
icatio
n
r
esu
lts
o
f
e
ac
h
C
NN
m
o
d
el.
1
,
2
,
3
,
4
,
an
d
5
in
th
e
eq
u
atio
n
a
r
e,
r
esp
ec
tiv
ely
,
t
h
e
v
o
tin
g
r
esu
lts
f
o
r
th
e
cl
ass
if
icatio
n
r
esu
lts
o
f
I
n
ce
p
tio
n
V3
,
Xce
p
tio
n
,
Den
s
N
et2
0
1
,
E
f
f
icie
n
tNetB
3
,
an
d
R
esNet5
0
at
e
p
o
ch
s
1
0
,
2
0
,
a
n
d
3
0
.
Me
a
n
wh
ile,
m
o
d
e
is
a
f
u
n
ctio
n
th
at
o
b
tain
s
th
e
m
ajo
r
ity
v
o
te
f
o
r
t
h
e
class
if
icatio
n
r
esu
lts
o
f
ea
ch
C
NN
m
o
d
el.
1
=
mode
(
ℎ
1
,
ℎ
2
,
ℎ
3
)
(
1
)
2
=
mode
(
ℎ
4
,
ℎ
5
,
ℎ
6
)
(
2
)
3
=
mode
(
ℎ
7
,
ℎ
8
,
ℎ
9
)
(
3
)
4
=
mod
e
(
ℎ
10
,
ℎ
11
,
ℎ
12
)
(
4
)
5
=
mode
(
ℎ
13
,
ℎ
14
,
ℎ
15
)
(
5
)
E
ac
h
b
ase
C
NN
m
o
d
el'
s
ar
ch
itectu
r
e
ca
n
o
b
tain
d
if
f
er
en
t
tu
m
o
r
class
if
icatio
n
r
esu
lts
.
T
h
e
p
o
ten
tial
f
o
r
d
if
f
er
en
t
class
if
icatio
n
r
esu
lts
is
d
u
e
to
th
e
o
th
er
lay
e
r
s
o
r
co
n
v
o
lu
tio
n
b
lo
ck
s
b
u
ilt
b
y
ea
ch
m
o
d
el.
T
h
e
in
v
o
lv
em
e
n
t
o
f
s
ev
er
al
C
NN
m
o
d
els
as
b
ase
m
o
d
els
with
t
h
eir
r
esp
ec
tiv
e
ad
v
a
n
tag
es
is
a
s
o
lu
tio
n
t
o
class
if
y
tu
m
o
r
ty
p
es
with
v
ar
ied
s
h
a
p
es,
s
izes,
o
r
p
o
s
itio
n
s
.
C
o
n
s
eq
u
en
tly
,
th
e
s
ec
o
n
d
s
tep
MV
is
ap
p
lied
to
th
e
r
esu
lts
o
f
th
e
f
ir
s
t
s
tep
MV
to
o
b
tain
t
h
e
f
in
al
class
if
icatio
n
r
esu
lts
th
at
co
n
s
id
er
all
th
e
c
lass
if
icatio
n
r
esu
lts
o
f
b
ase
C
NN
m
o
d
els.
Ma
th
em
atica
lly
,
th
e
two
-
s
tep
MV
ca
n
b
e
wr
itten
as (
6
)
.
=
mode
(
1
,
2
,
3
,
4
,
5
)
(
6
)
T
h
e
s
tep
s
in
Alg
o
r
ith
m
3
ar
e
u
s
ed
to
o
b
tain
th
e
f
in
al
class
if
icatio
n
r
esu
lts
with
th
e
p
r
o
p
o
s
ed
m
eth
o
d
(
two
-
s
tep
MV
)
.
T
h
e
alg
o
r
ith
m
'
s
in
p
u
t
is
all
b
ase
C
NN
m
o
d
els
th
at
h
av
e
u
n
d
er
g
o
n
e
a
lear
n
in
g
p
r
o
ce
s
s
b
ased
o
n
p
ar
am
eter
in
T
a
b
le
2
.
I
n
t
h
e
alg
o
r
ith
m
,
th
e
in
p
u
t
o
f
th
e
b
ase
C
NN
m
o
d
els
in
clu
d
es
M
11
,
M
12
,
M
13
wh
ich
s
h
o
w
th
e
lear
n
in
g
r
esu
lts
o
f
I
n
ce
p
tio
n
V3
,
M
21
,
M
22
,
M
23
ar
e
th
e
lear
n
in
g
r
esu
lts
o
f
Xce
p
tio
n
,
M
31
,
M
32
,
M
33
ar
e
th
e
lear
n
in
g
r
esu
lts
o
f
Den
s
N
et2
0
1
,
M
41
,
M
42
,
M
43
ar
e
th
e
lear
n
in
g
r
esu
lts
o
f
E
f
f
icien
tN
etB
3
,
an
d
M
51
,
M
51
,
M
51
ar
e
th
e
lear
n
in
g
r
esu
lts
o
f
R
esNet5
0
.
T
h
ese
lear
n
in
g
r
es
u
lts
ar
e
s
u
cc
es
s
iv
ely
ca
r
r
ied
o
u
t
at
ep
o
ch
s
1
0
,
2
0
,
an
d
3
0
f
o
r
ea
ch
m
o
d
el.
T
h
e
o
th
er
al
g
o
r
ith
m
in
p
u
t
is
X
,
w
h
ich
s
h
o
ws
th
e
b
r
ain
MRI
i
m
ag
e
o
f
th
e
test
in
g
d
ataset.
T
h
e
f
ir
s
t
s
te
p
o
f
th
e
al
g
o
r
ith
m
is
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
o
f
th
e
im
ag
e
to
o
b
tain
th
e
s
am
e
im
ag
e
s
ize
as
th
e
tr
ain
in
g
im
a
g
e
s
ize,
wh
ich
is
22
4×
2
2
4
×
3
.
T
h
e
n
e
x
t
s
t
ep
is
th
e
f
ir
s
t
class
if
icatio
n
o
f
tu
m
o
r
s
u
s
in
g
ea
ch
b
ase
C
NN
m
o
d
el
an
d
ep
o
c
h
with
th
e
i
n
p
u
t
im
a
g
e
(
Z
)
.
I
n
t
esti
n
g
th
e
two
-
s
tep
MV
m
eth
o
d
,
th
r
e
e
class
if
icatio
n
s
o
f
b
r
ain
tu
m
o
r
s
ar
e
ca
r
r
ied
o
u
t in
s
tag
es.
T
h
e
f
ir
s
t c
lass
if
icat
io
n
r
esu
lt is
s
h
o
wn
b
y
ℎ
,
wh
ich
is
d
eter
m
in
ed
b
y
th
e
ar
g
m
a
x
o
p
e
r
atio
n
o
n
th
e
s
o
f
tm
a
x
v
alu
e
o
f
ea
ch
b
ase
C
NN
m
o
d
el,
with
=
1
,
2
,
.
.
.
,
1
5
.
T
h
e
ℎ
v
alu
e
is
a
clas
s
lab
el
(
k
)
,
k
=0
if
class
if
ied
as
a
g
lio
m
a
tu
m
o
r
,
1
f
o
r
m
en
in
g
io
m
a,
2
f
o
r
n
o
tu
m
o
r
,
an
d
3
f
o
r
p
itu
itar
y
.
ℎ
in
clu
d
es
ℎ
1
,
ℎ
2
,
an
d
ℎ
3
wh
ich
ar
e
th
e
f
i
r
s
t
class
if
icatio
n
r
esu
lts
o
f
I
n
ce
p
tio
n
V
3
.
ℎ
4
ℎ
5
,
an
d
ℎ
6
ar
e
th
e
f
ir
s
t c
lass
if
icatio
n
r
esu
lts
o
f
Xce
p
tio
n
.
ℎ
7
,
ℎ
8
,
an
d
ℎ
9
ar
e
th
e
f
ir
s
t c
lass
if
icatio
n
r
esu
lts
o
f
Den
s
Net2
0
1
.
ℎ
10
,
ℎ
11
,
an
d
ℎ
12
ar
e
th
e
f
ir
s
t
class
if
icatio
n
r
e
s
u
lts
o
f
E
f
f
icien
tN
etB
3
.
ℎ
13
,
ℎ
14
,
an
d
ℎ
15
ar
e
th
e
f
ir
s
t
class
if
icatio
n
r
esu
lts
o
f
R
esN
et5
0
.
All
class
if
icatio
n
r
esu
lts
o
f
ea
ch
m
o
d
el
a
r
e
r
esp
ec
tiv
el
y
at
ep
o
c
h
s
1
0
,
2
0
,
an
d
3
0
.
T
h
e
in
itial
class
if
icati
o
n
r
esu
lts
at
ea
ch
ep
o
ch
an
d
t
h
e
C
NN
m
o
d
el
ar
e
th
en
f
o
r
w
ar
d
ed
to
th
e
s
ec
o
n
d
class
if
icatio
n
s
tag
e
u
s
in
g
th
e
f
ir
s
t
MV
(
o
n
e
-
s
tep
MV
)
with
th
e
r
esu
lts
in
d
icate
d
b
y
with
i
=
1
,
2
,
.
.
,
5
.
in
clu
d
es
1
,
2
,
3
,
4
,
an
d
5
,
wh
ich
ar
e
th
e
s
ec
o
n
d
class
if
icatio
n
r
esu
lts
with
o
n
e
-
s
tep
MV
u
s
in
g
(
1
)
,
(
2
)
,
(
3
)
,
(
4
)
,
a
n
d
(
5
)
.
T
h
e
r
esu
lts
o
f
th
e
s
ec
o
n
d
class
if
icatio
n
ar
e
th
en
f
o
r
war
d
ed
to
th
e
f
in
al
(
t
h
ir
d
)
class
if
icatio
n
s
tag
e
with
th
e
s
ec
o
n
d
MV
(
two
-
s
tep
MV
)
u
s
in
g
(
6
)
,
w
h
o
s
e
r
esu
lt
s
ar
e
in
d
icate
d
b
y
.
Fro
m
th
e
r
esu
lts
o
f
th
e
th
ir
d
class
if
icatio
n
,
it
ca
n
th
en
b
e
d
ec
id
ed
w
h
eth
er
th
e
MRI
im
ag
e
o
f
th
e
test
in
g
b
r
ain
is
class
if
ied
as
g
lio
m
a,
m
en
in
g
io
m
a
,
p
itu
itar
y
,
o
r
n
o
t
u
m
o
r
.
Alg
o
r
ith
m
3
,
in
its
im
p
lem
en
tatio
n
,
in
v
o
lv
es
th
e
class
if
icat
i
o
n
r
esu
lts
o
f
b
ase
C
NN
m
o
d
els
s
h
o
wn
in
Alg
o
r
ith
m
1
an
d
o
n
e
-
s
tep
M
V
in
Alg
o
r
ith
m
2
.
T
h
e
co
m
p
a
r
is
o
n
o
f
th
e
th
r
ee
alg
o
r
ith
m
s
c
an
b
e
s
ee
n
f
r
o
m
th
e
class
if
icatio
n
r
esu
lts
.
Alg
o
r
ith
m
3
is
a
co
n
tin
u
atio
n
o
f
Alg
o
r
ith
m
2
r
esu
lts
,
an
d
Alg
o
r
ith
m
2
is
a
co
n
tin
u
atio
n
o
f
Alg
o
r
ith
m
1
r
esu
lts
.
Alg
o
r
i
th
m
1
co
n
tain
s
th
e
s
tep
s
f
o
r
te
s
tin
g
MRI
im
ag
e
class
if
icat
io
n
u
s
in
g
th
e
tr
ain
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
Tw
o
-
s
tep
ma
jo
r
ity
vo
tin
g
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
…
(
I
r
w
a
n
B
u
d
i S
a
n
to
s
o
)
4093
r
esu
lts
o
f
ea
ch
C
NN
m
o
d
el
in
th
e
b
ase
m
o
d
els
at
ea
ch
ep
o
ch
.
At
th
e
s
am
e
tim
e,
th
e
h
y
p
er
p
ar
am
eter
tu
n
in
g
in
T
ab
le
2
is
th
e
tr
ea
tm
en
t
o
f
C
NN
m
o
d
el
p
a
r
am
eter
s
in
ten
d
e
d
f
o
r
th
e
lea
r
n
in
g
p
r
o
ce
s
s
to
p
r
o
v
id
e
g
o
o
d
m
o
d
el
tr
ain
in
g
r
esu
lts
.
Alg
o
r
ith
m
1
.
B
ase
C
NN
m
o
d
els
c
lass
if
icatio
n
I
n
p
u
t
:
M
11
,
M
12
,
M
13
{I
n
c
e
p
t
i
o
n
V3
},
M
21
, M
22
,
M
23
{
Xc
e
p
t
i
o
n
},
M
31
,
M
32
, M
33
{Den
s
N
e
t
2
0
1
},
M
41
, M
42
,
M
43
{E
f
f
i
c
i
e
n
t
N
e
t
B3
},
M
51
, M
51
,
M
51
{R
e
s
N
e
t
5
0
}{
T
h
e
t
r
a
i
n
i
n
g
r
e
su
l
t
s
o
f
e
a
c
h
m
o
d
e
l
a
re
c
o
n
se
c
u
t
i
v
e
l
y
a
t
e
p
o
c
h
s
1
0
,
2
0
,
a
n
d
3
0
,
r
e
sp
e
c
t
i
v
e
l
y
}
,
X
{B
r
a
i
n
M
RI
i
m
a
g
e
o
f
t
e
st
i
n
g
}
O
u
t
p
u
t
:
ℎ
{C
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
e
a
c
h
C
N
N
m
o
d
e
l
a
n
d
e
p
o
c
h
i
n
c
l
u
d
e
ℎ
1
,
ℎ
2
,
.
.
.
,
ℎ
15
}
1
Z
r
e
si
z
e
(
X,
2
2
4
,
2
2
4
,
3
)
{p
r
e
-
p
r
o
c
e
ssi
n
g
}
2
p
0
3
f
o
r
i
1
t
o
5
{
n
u
m
b
e
r
o
f
m
o
d
e
l
s}
4
f
o
r
i
1
t
o
3
{n
u
m
b
e
r
o
f
e
p
o
c
h
s {
1
0
,
2
0
,
3
0
}}
5
p
p
+
1
{i
n
d
e
x
o
f
c
l
a
ssi
f
i
c
a
t
i
o
n
}
6
ℎ
a
r
g
max
0≤
k
≤3
(
M
ij
(
Z
))
{
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
e
a
c
h
b
a
s
e
C
N
N
m
o
d
e
l
}
7
r
e
t
u
r
n
ℎ
Alg
o
r
ith
m
2
.
On
e
-
s
tep
m
ajo
r
ity
v
o
tin
g
I
n
p
u
t
:
M
11
,
M
12
,
M
13
{I
n
c
e
p
t
i
o
n
V3
},
M
21
, M
22
,
M
23
{
Xc
e
p
t
i
o
n
},
M
31
,
M
32
, M
33
{Den
s
N
e
t
2
0
1
},
M
41
, M
42
,
M
43
{E
f
f
i
c
i
e
n
t
N
e
t
B3
},
M
51
, M
51
,
M
51
{R
e
s
N
e
t
5
0
}{
T
h
e
t
r
a
i
n
i
n
g
r
e
su
l
t
s
o
f
e
a
c
h
m
o
d
e
l
a
re
c
o
n
se
c
u
t
i
v
e
l
y
a
t
e
p
o
c
h
s
1
0
,
2
0
,
a
n
d
3
0
,
r
e
sp
e
c
t
i
v
e
l
y
}
,
X
{B
r
a
i
n
M
RI
i
m
a
g
e
o
f
t
e
st
i
n
g
}
O
u
t
p
u
t
:
{C
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
o
n
e
-
s
t
e
p
m
a
j
o
ri
t
y
v
o
t
i
n
g
f
o
r
e
a
c
h
C
N
N
m
o
d
e
l
:
1
,
2
,
…
,
5
}
1
Z
r
e
si
z
e
(
X,
2
2
4
,
2
2
4
,
3
)
{p
r
e
-
p
r
o
c
e
ssi
n
g
}
2
p
0
3
f
o
r
i
1
t
o
5
{
n
u
m
b
e
r
o
f
m
o
d
e
l
s}
4
f
o
r
i
1
t
o
3
{n
u
m
b
e
r
o
f
e
p
o
c
h
s=
{
1
0
,
2
0
,
3
0
}}
5
p
p
+
1
{i
n
d
e
x
o
f
c
l
a
ssi
f
i
c
a
t
i
o
n
}
6
ℎ
a
r
g
max
0≤
k
≤3
(
M
ij
(
Z
))
{
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
e
a
c
h
b
a
s
e
C
N
N
m
o
d
e
l
}
7
f
o
r
i
1
t
o
5
{
n
u
m
b
e
r
o
f
m
o
d
e
l
s}
8
r
1
+
3
(
i
-
1
)
9
mo
d
e
(
ℎ
,
ℎ
+
1
,
ℎ
+
2
)
{t
h
e
f
i
rst
s
t
e
p
o
f
m
a
j
o
ri
t
y
v
o
t
i
n
g
}
10
r
e
t
u
r
n
Alg
o
r
ith
m
3
.
T
wo
-
s
tep
m
ajo
r
ity
v
o
tin
g
I
n
p
u
t
:
M
11
,
M
12
,
M
13
{I
n
c
e
p
t
i
o
n
V
3
},
M
21
,
M
22
,
M
23
{X
c
e
p
t
i
o
n
}
,
M
31
,
M
32
,
M
33
{De
n
sN
e
t
2
0
1
},
M
41
,
M
42
,
M
43
{
Ef
f
i
c
i
e
n
t
N
e
t
B
3
},
M
51
,
M
51
,
M
51
{R
e
sN
e
t
5
0
}{
t
h
e
t
ra
i
n
i
n
g
r
e
su
l
t
s
o
f
e
a
c
h
m
o
d
e
l
a
r
e
c
o
n
s
e
c
u
t
i
v
e
l
y
a
t
e
p
o
c
h
s
1
0
,
2
0
,
a
n
d
3
0
,
r
e
sp
e
c
t
i
v
e
l
y
},
X
{
Bra
i
n
MR
I
i
m
a
g
e
o
f
t
e
st
i
n
g
}
O
u
t
p
u
t
:
{C
l
a
ss
i
f
i
c
a
t
i
o
n
r
e
su
l
t
}
1
Z
r
e
si
z
e
(
X,
2
2
4
,
2
2
4
,
3
)
{p
r
e
-
p
r
o
c
e
ssi
n
g
}
2
p
0
3
f
o
r
i
1
t
o
5
{
n
u
m
b
e
r
o
f
m
o
d
e
l
s}
4
f
o
r
i
1
t
o
3
{n
u
m
b
e
r
o
f
e
p
o
c
h
s=
{
1
0
,
2
0
,
3
0
}}
5
p
p
+
1
{i
n
d
e
x
o
f
c
l
a
ssi
f
i
c
a
t
i
o
n
}
6
ℎ
a
r
g
max
0≤
k
≤3
(
M
ij
(
Z
))
{c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
b
a
se
C
N
N
m
o
d
e
l
,
k
=
0
(
g
l
i
o
m
a
)
,
k
=
1
(
m
e
n
i
n
g
i
o
m
a
)
,
k
=
2
(
n
o
t
u
m
o
r)
,
k
=
3
(
p
i
t
u
i
t
a
ry)}
7
f
o
r
i
1
t
o
5
{
n
u
m
b
e
r
o
f
m
o
d
e
l
s}
8
r
1
+
3
(
i
-
1
)
{i
n
d
e
x
o
f
c
l
a
ssi
f
i
c
a
t
i
o
n
}
9
mo
d
e
(
ℎ
,
ℎ
+
1
,
ℎ
+
2
)
{t
h
e
f
i
rst
s
t
e
p
o
f
m
a
j
o
ri
t
y
v
o
t
i
n
g
}
10
m
o
d
e
(
)
)
{t
h
e
t
w
o
st
e
p
o
f
m
a
j
o
ri
t
y
v
o
t
i
n
g
}
11
r
e
t
u
r
n
2
.
5
.
P
er
f
o
r
m
a
nce
ev
a
lua
t
io
n o
f
m
e
t
ho
ds
I
n
th
is
s
tu
d
y
,
to
ev
alu
ate
th
e
m
eth
o
d
'
s
p
er
f
o
r
m
an
ce
in
b
r
ain
tu
m
o
r
class
if
icatio
n
,
s
ev
er
al
in
d
icato
r
m
ea
s
u
r
es
wer
e
u
s
ed
,
t
h
ey
we
r
e
ac
cu
r
ac
y
(
)
,
p
r
ec
is
io
n
(
)
,
s
en
s
itiv
ity
(
)
,
s
p
ec
if
icity
(
)
,
an
d
F
-
s
co
r
e
(
)
[
3
9
]
.
All
o
f
t
h
ese
in
d
icato
r
s
ar
e
o
b
tain
e
d
b
ased
o
n
th
e
v
al
u
es
o
f
tr
u
e
p
o
s
itiv
e
(
)
,
f
alse
n
eg
ativ
e
(
)
,
tr
u
e
n
e
g
ativ
e
(
)
,
an
d
f
alse
p
o
s
itiv
e
(
)
.
T
h
ese
p
er
f
o
r
m
a
n
ce
in
d
icato
r
s
ar
e
d
ef
in
e
d
f
o
r
ea
ch
d
a
taset
la
b
el,
n
am
ely
g
lio
m
a,
m
en
in
g
io
m
a,
p
itu
itar
y
,
an
d
n
o
tu
m
o
r
.
Fo
r
g
lio
m
a,
is
th
e
n
u
m
b
er
o
f
tim
es
th
e
g
lio
m
a
b
r
ain
MRI
im
ag
e,
b
ased
o
n
t
h
e
class
if
icatio
n
r
esu
lts
,
is
lab
eled
as
g
lio
m
a,
is
th
e
n
u
m
b
er
o
f
tim
es
th
e
g
lio
m
a
b
r
ain
MRI
im
ag
e,
b
a
s
ed
o
n
th
e
class
if
icatio
n
r
esu
lts
,
is
lab
eled
as
o
th
er
th
an
g
lio
m
a,
is
th
e
n
u
m
b
er
o
f
tim
es
th
e
b
r
ain
M
R
I
im
ag
e
o
th
er
th
an
g
lio
m
a
is
lab
eled
as
o
th
er
th
an
g
lio
m
a,
an
d
is
th
e
n
u
m
b
er
o
f
tim
es
th
e
b
r
ain
M
R
I
im
ag
e
o
th
er
th
an
g
li
o
m
a
is
lab
eled
as
g
lio
m
a
in
t
h
e
s
am
e
way
.
Me
an
wh
ile,
,
,
,
,
a
n
d
ar
e
d
eter
m
in
ed
f
o
r
ea
ch
lab
el
a
n
d
ar
e
r
esp
ec
tiv
e
ly
d
ef
i
n
ed
in
(
7
)
,
(
8
)
,
(
9
)
,
(
1
0
)
,
an
d
(
1
1
)
.
=
(
+
)
/
(
+
+
+
)
(
7
)
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
.
4
,
Au
g
u
s
t
20
25
:
4
0
8
7
-
4098
4094
=
/
(
+
)
(
8
)
=
/
(
+
)
(
9
)
=
/
(
+
)
(
1
0
)
=
2
(
)
(
)
/
(
+
)
(
1
1
)
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
im
p
lem
en
tatio
n
o
f
all
th
e
m
eth
o
d
s
u
s
ed
Go
o
g
le
C
o
lab
in
th
e
lear
n
in
g
p
r
o
ce
s
s
o
f
ea
ch
b
ase
C
NN
m
o
d
el,
t
h
e
p
er
f
o
r
m
an
ce
test
in
g
o
f
b
ase
C
NN
m
o
d
els,
a
n
d
t
h
e
p
e
r
f
o
r
m
an
ce
test
in
g
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
(
two
-
s
tep
MV
)
in
b
r
ain
tu
m
o
r
class
if
icatio
n
.
T
ab
le
3
s
h
o
w
s
th
e
r
esu
lts
o
f
tes
tin
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
b
ase
C
NN
m
o
d
els
an
d
two
-
s
tep
M
V
in
b
r
ain
t
u
m
o
r
class
if
icatio
n
o
n
t
h
e
n
ick
p
ar
v
a
r
d
ataset.
T
h
e
r
esu
lts
o
f
test
in
g
th
e
b
ase
C
NN
m
o
d
els
o
n
th
e
test
in
g
d
ataset
f
o
r
th
e
I
n
ce
p
tio
n
V3
m
o
d
el
p
r
o
d
u
ce
d
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
8
.
7
8
%
at
ep
o
c
h
,
9
9
.
2
4
%
at
ep
o
ch
2
0
,
an
d
9
9
.
0
8
%
at
ep
o
ch
3
0
.
Fo
r
Xce
p
tio
n
,
it
o
b
tain
ed
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
9
.
2
4
% a
t e
p
o
ch
1
0
,
9
9
.
4
7
% a
t e
p
o
ch
2
0
,
a
n
d
9
9
.
3
9
% a
t e
p
o
ch
3
0
.
Den
s
Net2
0
1
o
b
tain
e
d
a
tu
m
o
r
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
9
.
1
6
%
at
ep
o
c
h
1
0
,
9
9
.
3
6
%
at
ep
o
ch
2
0
,
a
n
d
9
9
.
3
9
%
at
ep
o
ch
3
0
.
On
th
e
o
th
er
h
an
d
,
E
f
f
icien
tNetB
3
,
at
ep
o
c
h
s
1
0
,
2
0
,
a
n
d
3
0
,
p
r
o
d
u
ce
d
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
9
.
4
7
%,
9
9
.
0
8
%,
an
d
9
9
.
3
1
%.
Me
an
w
h
ile,
R
esNet5
0
at
ep
o
ch
s
1
0
,
2
0
,
a
n
d
3
0
o
b
t
ain
ed
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
8
.
2
5
%,
9
8
.
8
6
%,
an
d
9
8
.
7
0
%.
Fro
m
th
ese
r
esu
lts
,
Xce
p
tio
n
at
ep
o
ch
2
0
an
d
E
f
f
icien
tNetB
3
at
ep
o
ch
1
0
y
ield
ed
th
e
h
ig
h
est
ac
cu
r
ac
y
am
o
n
g
th
e
b
ase
C
NN
m
o
d
els
an
d
th
e
b
est
av
er
a
g
e
s
en
s
itiv
ity
,
p
r
ec
is
io
n
,
s
p
ec
if
icity
,
an
d
F
-
s
co
r
e
co
m
p
ar
ed
to
o
th
er
m
o
d
els.
T
h
e
lo
west
ac
cu
r
ac
y
o
f
th
e
b
ase
C
NN
m
o
d
el
is
R
es
Net5
0
at
ep
o
ch
1
0
,
as
well
a
s
th
e
lo
west a
v
er
ag
e
s
en
s
itiv
ity
,
p
r
ec
is
io
n
,
s
p
ec
if
icity
,
a
n
d
F
-
S
co
r
e
am
o
n
g
th
e
o
th
er
b
ase
C
NN
m
o
d
els.
T
ab
le
3
.
Per
f
o
r
m
an
ce
o
f
b
ase
C
NN
m
o
d
els an
d
two
-
s
tep
m
a
jo
r
ity
v
o
tin
g
M
o
d
e
l
Ac
c
u
r
a
c
y
A
v
e
r
a
g
e
p
e
r
f
o
r
ma
n
c
e
P
r
e
c
i
s
i
o
n
S
e
n
s
i
t
i
v
y
S
p
e
c
i
f
i
c
i
t
y
F
-
sco
r
e
B
a
se
C
N
N
m
o
d
e
l
s
(
e
p
o
c
h
)
(
1
)
I
n
c
e
p
t
i
o
n
V
3
(
1
0
)
0
.
9
8
7
8
0
.
9
8
6
8
0
.
9
8
6
8
0
.
9
9
6
0
0
.
9
8
6
8
(
2
)
I
n
c
e
p
t
i
o
n
V
3
(
2
0
)
0
.
9
9
2
4
0
.
9
9
1
8
0
.
9
9
1
7
0
.
9
9
7
5
0
.
9
9
1
7
(
3
)
I
n
c
e
p
t
i
o
n
V
3
(
3
0
)
0
.
9
9
0
8
0
.
9
9
0
2
0
.
9
9
0
7
0
.
9
9
7
0
0
.
9
9
0
4
(
4
)
X
c
e
p
t
i
o
n
(
1
0
)
0
.
9
9
2
4
0
.
9
9
2
0
0
.
9
9
1
7
0
.
9
9
7
5
0
.
9
9
1
9
(
5
)
X
c
e
p
t
i
o
n
(
2
0
)
0
.
9
9
4
7
0
.
9
9
4
3
0
.
9
9
4
2
0
.
9
9
8
3
0
.
9
9
4
2
(
6
)
X
c
e
p
t
i
o
n
(
3
0
)
0
.
9
9
3
9
0
.
9
9
3
5
0
.
9
9
3
8
0
.
9
9
8
0
0
.
9
9
3
6
(
7
)
D
e
n
sN
e
t
2
0
1
(
1
0
)
0
.
9
9
1
6
0
.
9
9
1
0
0
.
9
9
1
3
0
.
9
9
7
3
0
.
9
9
1
2
(
8
)
D
e
n
sN
e
t
2
0
1
(
2
0
)
0
.
9
9
3
9
0
.
9
9
3
6
0
.
9
9
3
4
0
.
9
9
8
0
0
.
9
9
3
5
(
9
)
D
e
n
sN
e
t
2
0
1
(
3
0
)
0
.
9
9
3
9
0
.
9
9
3
4
0
.
9
9
3
4
0
.
9
9
8
0
0
.
9
9
3
4
(
1
0
)
Ef
f
i
c
i
e
n
t
N
e
t
B
3
(
1
0
)
0
.
9
9
4
7
0
.
9
9
4
2
0
.
9
9
4
2
0
.
9
9
8
3
0
.
9
9
4
2
(
1
1
)
Ef
f
i
c
i
e
n
t
N
e
t
B
3
(
2
0
)
0
.
9
9
0
8
0
.
9
9
0
6
0
.
9
9
0
5
0
.
9
9
7
0
0
.
9
9
0
5
(
1
2
)
Ef
f
i
c
i
e
n
t
N
e
t
B
3
(
3
0
)
0
.
9
9
3
1
0
.
9
9
3
0
0
.
9
9
2
6
0
.
9
9
7
7
0
.
9
9
2
8
(
1
3
)
R
e
sN
e
t
5
0
(
1
0
)
0
.
9
8
2
5
0
.
9
8
1
7
0
.
9
8
0
9
0
.
9
9
4
3
0
.
9
8
1
1
(
1
4
)
R
e
sN
e
t
5
0
(
2
0
)
0
.
9
8
8
6
0
.
9
8
8
0
0
.
9
8
7
6
0
.
9
9
6
2
0
.
9
8
7
8
(
1
5
)
R
e
sN
e
t
5
0
(
3
0
)
0
.
9
8
7
0
0
.
9
8
6
4
0
.
9
8
5
9
0
.
9
9
5
8
0
.
9
8
6
1
O
n
e
-
s
t
e
p
M
V
(
1
6
)
M
V
(
1
,
2
,
3
)
0
.
9
9
4
7
0
.
9
9
4
3
0
.
9
9
4
2
0
.
9
9
8
3
0
.
9
9
4
2
(
1
7
)
M
V
(
4
,
5
,
6
)
0
.
9
9
3
9
0
.
9
9
3
5
0
.
9
9
3
8
0
.
9
9
8
0
0
.
9
9
3
6
(
1
8
)
M
V
(
7
,
8
,
9
)
0
.
9
9
5
4
0
.
9
9
5
1
0
.
9
9
5
0
0
.
9
9
8
5
0
.
9
9
5
0
(
1
9
)
M
V
(
1
0
,
1
1
,
1
2
)
0
.
9
9
6
2
0
.
9
9
5
9
0
.
9
9
5
8
0
.
9
9
8
8
0
.
9
9
5
9
(
2
0
)
M
V
(
1
3
,
1
4
,
1
5
)
0
.
9
8
8
6
0
.
9
8
7
9
0
.
9
8
7
6
0
.
9
9
6
2
0
.
9
8
7
7
Tw
o
-
s
t
e
p
M
V
(
2
1
)
M
V
(
1
6
,
1
7
,
1
8
,
1
9
,
2
0
)
0
.
9
9
6
9
0
.
9
9
6
7
0
.
9
9
6
7
0
.
9
9
9
0
0
.
9
9
6
7
T
h
e
test
r
esu
lts
f
o
r
th
e
f
ir
s
t
s
t
ep
MV
o
n
ea
ch
C
NN
m
o
d
el
with
d
if
f
er
en
t
ep
o
c
h
s
,
th
e
m
aj
o
r
ity
v
o
tin
g
o
n
th
e
class
if
icatio
n
r
esu
lts
o
f
th
e
E
f
f
icien
tNetB
3
m
o
d
el
(
MV
(
1
0
,
1
1
,
1
2
)
)
,
g
av
e
an
ac
c
u
r
ac
y
o
f
9
9
.
6
2
%
an
d
wer
e
th
e
b
est
am
o
n
g
th
e
f
ir
s
t
s
tep
MV
s
o
n
o
th
er
m
o
d
els.
T
h
e
lo
west
MV
ac
cu
r
ac
y
am
o
n
g
th
e
o
th
e
r
m
o
d
els
was
th
e
MV
o
n
th
e
R
esNet5
0
class
if
icatio
n
r
esu
lts
at
d
if
f
er
en
t
ep
o
ch
s
(
MV
(
1
3
,
1
4
,
1
5
)
)
,
with
an
ac
cu
r
ac
y
o
f
98.
8
6
%.
Me
an
w
h
ile,
th
e
s
ec
o
n
d
s
tep
MV
(
two
-
s
tep
MV
)
g
av
e
b
etter
class
if
icatio
n
p
er
f
o
r
m
an
ce
th
an
th
e
f
ir
s
t
s
tep
MV
an
d
o
n
all
b
ase
C
NN
m
o
d
els.
Fro
m
th
ese
r
esu
lts
,
two
-
s
tep
MV
(
MV
(
1
6
,
1
7
,
1
8
,
1
9
,
2
0
)
)
in
cr
ea
s
ed
class
if
icatio
n
ac
cu
r
ac
y
b
y
0
.
2
2
%
to
1
.
4
4
%
o
n
b
ase
C
NN
m
o
d
els
an
d
0
.
0
7
%
t
o
0
.
8
3
%
m
a
jo
r
ity
v
o
tin
g
o
n
th
e
f
ir
s
t
s
tep
.
T
wo
-
s
tep
MV
im
p
r
o
v
es
class
if
icatio
n
p
er
f
o
r
m
a
n
c
e
o
n
b
ase
C
NN
m
o
d
els
th
r
o
u
g
h
two
s
tep
s
o
f
MV
.
T
h
e
f
ir
s
t
s
tep
o
f
MV
will
s
t
r
en
g
th
en
th
e
class
if
icatio
n
p
e
r
f
o
r
m
a
n
ce
o
f
ea
c
h
b
ase
C
NN
m
o
d
el.
T
h
e
MV
cla
s
s
if
icatio
n
r
esu
lts
o
f
ea
c
h
C
NN
m
o
d
el
im
p
lem
en
te
d
b
y
lear
n
in
g
with
d
if
f
er
e
n
t
h
y
p
er
p
a
r
am
eter
s
an
d
ep
o
ch
s
ca
n
r
ed
u
ce
th
e
u
n
ce
r
ta
in
ty
(
co
in
ci
d
en
ce
f
ac
to
r
)
o
f
t
h
e
class
if
icatio
n
r
esu
lts
o
f
ea
ch
m
o
d
el.
T
wo
-
s
tep
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
Tw
o
-
s
tep
ma
jo
r
ity
vo
tin
g
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
…
(
I
r
w
a
n
B
u
d
i S
a
n
to
s
o
)
4095
MV
ap
p
lied
to
o
n
e
-
s
tep
MV
will
s
tr
en
g
th
en
th
e
clas
s
if
icat
io
n
o
f
s
p
ec
if
ic
tu
m
o
r
ty
p
es
a
n
d
leav
e
th
e
wea
k
C
NN
m
o
d
el
in
th
e
class
if
icatio
n
o
f
th
at
tu
m
o
r
ty
p
e.
T
ab
le
4
s
h
o
ws
th
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
a
g
ain
s
t
th
e
b
ase
C
NN
m
o
d
el
an
d
th
e
b
est
o
n
e
-
s
tep
MV
in
m
o
r
e
d
e
p
th
.
On
e
-
s
tep
MV
y
ie
ld
ed
th
e
b
est p
e
r
f
o
r
m
an
ce
t
o
th
e
E
f
f
icien
tNetB
3
m
o
d
el
at
d
if
f
er
en
t e
p
o
ch
s
,
a
n
d
th
e
two
-
s
tep
MV
im
p
r
o
v
ed
th
e
p
r
ec
is
io
n
o
f
g
lio
m
a
an
d
m
en
i
n
g
io
m
a
class
if
icatio
n
f
o
r
th
e
o
n
e
-
s
tep
MV
b
y
0
.
0
1
%
a
n
d
0
.
3
2
%,
r
esp
ec
tiv
ely
.
T
h
e
two
-
s
tep
M
V
also
p
r
o
v
id
e
d
im
p
r
o
v
e
m
en
ts
i
n
th
e
s
en
s
itiv
ity
o
f
g
lio
m
a
class
if
icatio
n
b
y
0
.
3
3
%,
im
p
r
o
v
e
m
en
ts
in
th
e
s
p
ec
if
icity
o
f
m
en
in
g
io
m
a
clas
s
if
icatio
n
b
y
0
.
1
0
%,
a
n
d
a
u
to
m
atica
lly
p
r
o
v
i
d
ed
im
p
r
o
v
em
e
n
ts
in
t
h
e
F
-
s
co
r
e.
I
n
ad
d
itio
n
,
two
-
s
tep
MV
al
s
o
im
p
r
o
v
ed
p
er
f
o
r
m
a
n
ce
ag
ain
s
t
th
e
b
est
b
ase
C
NN
m
o
d
el
(
Xce
p
tio
n
at
ep
o
ch
2
0
an
d
E
f
f
icien
tNetB
3
at
e
p
o
ch
1
0
)
,
n
am
ely
g
lio
m
a
class
if
icatio
n
p
r
ec
is
io
n
o
f
0
.
3
4
%
to
0
.
6
7
%,
m
en
in
g
i
o
m
a
class
if
icatio
n
p
r
ec
is
io
n
o
f
0
.
3
3
%
to
0
.
6
5
%,
g
lio
m
a
class
if
icatio
n
s
en
s
itiv
ity
o
f
0
.
3
3
%
to
0
.
6
6
%,
m
en
in
g
io
m
a
class
if
icatio
n
s
en
s
it
iv
ity
o
f
0
.
3
2
%
to
0
.
6
5
%,
g
lio
m
a
class
if
icatio
n
s
p
ec
if
icity
o
f
0
.
1
%
to
0
.
2
%,
m
en
in
g
io
m
a
class
if
icatio
n
s
p
ec
if
icity
o
f
0
.
1
0
%
to
0
.
2
%,
g
lio
m
a
class
i
f
icatio
n
F
-
s
co
r
e
o
f
0
.
5
%,
an
d
m
en
in
g
io
m
a
class
if
icatio
n
F
-
s
co
r
e
o
f
0
.
4
9
%.
T
ab
le
4
.
Dee
p
p
er
f
o
r
m
an
ce
o
f
th
e
b
est b
ase
C
NN
m
o
d
els,
o
n
e
-
s
tep
MV
,
an
d
two
-
s
tep
MV
M
o
d
e
l
La
b
e
l
A
v
e
r
a
g
e
P
e
r
f
o
r
ma
n
c
e
P
r
e
c
i
s
i
o
n
S
e
n
s
i
t
i
v
y
S
p
e
c
i
f
i
c
i
t
y
F
-
S
c
o
r
e
B
a
se
C
N
N
m
o
d
e
l
s
(
e
p
o
c
h
)
X
c
e
p
t
i
o
n
(
2
0
)
g
l
i
o
m
a
0.
9
9
3
3
0.
9
8
6
7
0.
9
9
8
0
0.
9
9
0
0
men
i
n
g
i
o
m
a
0.
9
8
3
8
0.
9
9
3
5
0.
9
9
5
0
0.
9
8
8
6
n
o
t
u
m
o
r
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
p
i
t
u
i
t
a
r
y
1
.
0
0
0
0
0.
9
9
6
7
1
.
0
0
0
0
0.
9
9
8
3
Ef
f
i
c
i
e
n
t
N
e
t
B
3
(
1
0
)
g
l
i
o
m
a
0.
9
9
0
0
0.
9
9
0
0
0.
9
9
7
0
0.
9
9
0
0
men
i
n
g
i
o
m
a
0.
9
8
7
0
0.
9
9
0
2
0.
9
9
6
0
0.
9
8
8
6
n
o
t
u
m
o
r
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
p
i
t
u
i
t
a
r
y
1
.
0
0
0
0
0.
9
9
.
6
7
1
.
0
0
0
0
0.
9
9
8
3
O
n
e
-
s
t
e
p
M
V
MV
(
1
0
,
1
1
,
1
2
)
g
l
i
o
m
a
0.
9
9
6
6
0.
9
9
0
0
0.
9
9
9
0
0.
9
9
3
3
men
i
n
g
i
o
m
a
0.
9
8
7
1
0.
9
9
6
7
0.
9
9
6
0
0.
9
9
1
9
n
o
t
u
m
o
r
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
p
i
t
u
i
t
a
r
y
1
.
0
0
0
0
0.
9
9
6
7
1
.
0
0
0
0
0.
9
9
.
8
3
Tw
o
-
s
t
e
p
M
V
MV
(
1
6
,
1
7
,
1
8
,
1
9
,
2
0
)
g
l
i
o
m
a
0.
9
9
.
6
7
0.
9
9
3
3
0.
9
9
9
0
0.
9
9
5
0
men
i
n
g
i
o
m
a
0.
9
9
.
0
3
0.
9
9
6
7
0.
9
9
7
0
0.
9
9
3
5
n
o
t
u
m
o
r
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
1
.
0
0
0
0
p
i
t
u
i
t
a
r
y
1
.
0
0
0
0
0.
9
9
6
7
1
.
0
0
0
0
0.
9
9
8
3
4.
CO
M
P
ARI
SO
N
WI
T
H
E
X
I
ST
I
NG
M
E
T
H
O
DS
T
h
e
co
m
p
ar
ativ
e
r
esu
lts
o
f
e
v
alu
atin
g
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
with
s
ev
e
r
al
p
r
ev
io
u
s
m
eth
o
d
s
o
n
th
e
s
am
e
test
in
g
d
ataset
ar
e
in
v
e
s
tig
atio
n
ev
id
en
ce
o
f
t
h
e
m
et
h
o
d
s
in
tu
m
o
r
class
if
icatio
n
.
Ap
p
ly
in
g
two
-
s
tep
m
ajo
r
ity
v
o
tin
g
to
th
e
b
ase
C
NN
m
o
d
els
(
th
e
two
-
s
tep
MV
)
p
r
o
d
u
ce
s
b
etter
p
er
f
o
r
m
an
ce
th
an
th
e
ex
is
tin
g
m
eth
o
d
s
.
T
a
b
le
5
s
h
o
ws
th
at
th
e
p
r
o
p
o
s
ed
m
eth
o
d
is
b
etter
th
an
th
e
ex
is
tin
g
m
et
h
o
d
s
,
w
ith
a
d
if
f
e
r
en
ce
o
f
1
.
8
5
%
to
7
.
6
7
%.
T
h
e
i
n
v
o
l
v
em
en
t
o
f
s
ev
er
al
b
ase
C
N
N
m
o
d
els
in
two
-
s
tep
MV
s
tr
en
g
th
en
s
t
u
m
o
r
class
if
icatio
n
b
y
ab
an
d
o
n
in
g
wea
k
C
NN
m
o
d
els.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
ca
n
im
p
r
o
v
e
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
o
f
b
r
ain
t
u
m
o
r
t
y
p
es,
esp
ec
ially
g
lio
m
a
an
d
m
e
n
in
g
io
m
a
tu
m
o
r
s
th
at
ar
e
alm
o
s
t
th
e
s
am
e
s
h
ap
e
an
d
s
ize.
T
ab
le
5
.
Per
f
o
r
m
an
ce
co
m
p
a
r
is
o
n
o
f
two
-
s
tep
m
aj
o
r
ity
v
o
tin
g
with
th
e
ex
is
tin
g
m
eth
o
d
s
R
e
f
e
r
e
n
c
e
s
M
e
t
h
o
d
s
A
c
c
u
r
a
c
y
R
a
s
h
e
e
d
e
t
a
l
.
[
4
0
]
I
mag
e
e
n
h
a
n
c
e
me
n
t
a
n
d
C
N
N
0
.
9
7
8
4
S
h
i
l
a
s
k
a
r
e
t
a
l
.
[
4
1
]
HOG
a
n
d
X
G
B
o
o
st
0
.
9
2
0
2
A
t
h
a
a
n
d
C
h
a
k
i
[
4
2
]
S
S
B
TC
N
e
t
0
.
9
6
5
P
r
o
p
o
se
d
M
e
t
h
o
d
s
Tw
o
-
st
e
p
ma
j
o
r
i
t
y
v
o
t
i
n
g
0
.
9
9
6
9
5.
CO
NCLU
SI
O
N
T
wo
-
s
tep
m
ajo
r
ity
v
o
tin
g
is
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
in
th
is
s
tu
d
y
th
at
a
p
p
lies
two
-
s
tep
m
ajo
r
ity
v
o
tin
g
to
th
e
r
esu
lts
o
f
th
e
class
if
icat
io
n
o
f
b
ase
C
NN
m
o
d
els
b
ased
o
n
b
r
ai
n
MRI
im
ag
es.
T
h
e
f
ir
s
t
s
tep
is
m
ajo
r
ity
v
o
tin
g
o
n
th
e
r
esu
lts
o
f
tu
m
o
r
class
if
icatio
n
o
n
ea
ch
b
ase
C
NN
m
o
d
el
with
d
if
f
er
en
t
e
p
o
ch
s
(
1
0
,
2
0
,
an
d
3
0
ep
o
ch
s
)
(
o
n
e
-
s
tep
m
ajo
r
ity
v
o
tin
g
)
.
T
h
e
b
ase
C
NN
m
o
d
els
in
clu
d
ed
I
n
ce
p
tio
n
V3
,
Xce
p
tio
n
,
Den
s
Net2
0
1
,
E
f
f
icien
tNetB
3
,
an
d
R
esNet5
0
.
T
h
e
s
ec
o
n
d
s
tep
is
m
ajo
r
ity
v
o
tin
g
o
n
t
h
e
r
esu
lts
o
f
all
t
h
e
m
ajo
r
ity
v
o
tin
g
i
n
th
e
f
ir
s
t
s
tep
.
T
h
e
test
u
s
ed
t
h
e
N
ick
p
ar
v
ar
d
ataset.
Ou
r
p
r
o
p
o
s
ed
m
eth
o
d
p
r
o
d
u
ce
d
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
p
r
ec
is
io
n
,
s
p
ec
if
icity
,
an
d
F
-
s
co
r
e
in
tu
m
o
r
class
if
icatio
n
o
f
9
9
.
6
9
%,
9
9
.
6
7
%,
9
9
.
6
7
%,
9
9
.
9
0
%,
an
d
9
9
.
6
7
%,
r
esp
ec
tiv
ely
.
T
wo
-
s
tep
m
ajo
r
i
ty
v
o
tin
g
in
c
r
ea
s
ed
ac
c
u
r
ac
y
o
v
er
b
ase
C
NN
m
o
d
els
a
n
d
th
e
o
n
e
-
s
tep
m
ajo
r
ity
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
.
4
,
Au
g
u
s
t
20
25
:
4
0
8
7
-
4098
4096
v
o
tin
g
.
T
h
e
d
etailed
p
er
f
o
r
m
an
ce
im
p
r
o
v
e
m
en
t
o
cc
u
r
r
ed
in
class
if
y
in
g
g
lio
m
a
an
d
m
en
in
g
io
m
a
tu
m
o
r
s
,
wh
ich
wer
e
alm
o
s
t
s
im
ilar
i
n
s
h
ap
e
an
d
s
ize.
I
n
ad
d
itio
n
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
was
b
etter
th
an
s
ev
er
al
ex
is
tin
g
m
eth
o
d
s
in
th
e
test
in
g
u
s
in
g
th
e
s
am
e
d
ata
s
et.
T
h
er
ef
o
r
e,
o
u
r
p
r
o
p
o
s
ed
m
eth
o
d
ca
n
h
elp
n
eu
r
o
lo
g
is
ts
in
tu
m
o
r
class
if
icatio
n
.
Ad
d
i
tio
n
ally
,
f
o
r
clin
ical
im
p
lem
en
tatio
n
,
th
er
e
is
a
p
o
ten
tial
f
o
r
im
p
r
o
v
in
g
t
h
e
p
er
f
o
r
m
an
ce
o
f
t
u
m
o
r
class
if
icatio
n
b
ased
o
n
b
r
ain
MRI
i
m
ag
es
b
y
ad
d
i
n
g
tr
ai
n
in
g
o
r
test
in
g
MRI
im
ag
e
d
ata.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
s
tu
d
y
is
en
tire
ly
f
u
n
d
ed
b
y
Un
iv
er
s
itas
I
s
lam
Ne
g
er
i
Ma
u
lan
a
Ma
lik
I
b
r
ah
i
m
Ma
lan
g
,
I
n
d
o
n
esia,
th
r
o
u
g
h
th
e
Nati
o
n
al
Dev
elo
p
m
e
n
t
Ap
p
lied
R
esear
ch
(
Pen
elitian
T
er
ap
an
Pen
g
em
b
an
g
a
n
Nasio
n
al)
s
ch
em
e
with
co
n
tr
a
ct
n
u
m
b
er
1
1
4
3
A/L
P2
M/T
L
.
0
0
/0
2
/2
0
2
4
.
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
t
r
ib
u
to
r
R
o
les
T
a
x
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
i
d
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
I
r
wan
B
u
d
i San
to
s
o
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Sh
o
f
f
in
Nah
wa
Utam
a
✓
✓
✓
✓
✓
✓
Su
p
r
iy
o
n
o
✓
✓
✓
✓
✓
✓
✓
✓
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
h
e
d
ata
th
at
s
u
p
p
o
r
t
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
o
p
e
n
ly
av
ailab
le
at
h
ttp
s
://www.
k
ag
g
le.
co
m
/d
atasets
/m
aso
u
d
n
ick
p
ar
v
a
r
/b
r
ain
-
tu
m
o
r
-
mri
-
d
ataset,
r
ef
er
e
n
ce
n
u
m
b
er
[
2
4
]
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
P
e
r
e
i
r
a
,
A
.
P
i
n
t
o
,
V
.
A
l
v
e
s,
a
n
d
C
.
A
.
S
i
l
v
a
,
“
B
r
a
i
n
t
u
mo
r
s
e
g
m
e
n
t
a
t
i
o
n
u
si
n
g
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
i
n
M
R
I
i
ma
g
e
s
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
M
e
d
i
c
a
l
I
m
a
g
i
n
g
,
v
o
l
.
3
5
,
n
o
.
5
,
p
p
.
1
2
4
0
–
1
2
5
1
,
M
a
y
2
0
1
6
,
d
o
i
:
1
0
.
1
1
0
9
/
T
M
I
.
2
0
1
6
.
2
5
3
8
4
6
5
.
[
2
]
M
.
S
.
I
.
K
h
a
n
e
t
a
l
.
,
“
A
c
c
u
r
a
t
e
b
r
a
i
n
t
u
m
o
r
d
e
t
e
c
t
i
o
n
u
si
n
g
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
C
o
m
p
u
t
a
t
i
o
n
a
l
a
n
d
S
t
r
u
c
t
u
r
a
l
Bi
o
t
e
c
h
n
o
l
o
g
y
J
o
u
r
n
a
l
,
v
o
l
.
2
0
,
p
p
.
4
7
3
3
–
4
7
4
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
s
b
j
.
2
0
2
2
.
0
8
.
0
3
9
.
[
3
]
A
.
R
.
L
o
u
g
h
a
n
e
t
a
l
.
,
“
D
e
a
t
h
-
r
e
l
a
t
e
d
d
i
s
t
r
e
ss
i
n
a
d
u
l
t
p
r
i
m
a
r
y
b
r
a
i
n
t
u
m
o
r
p
a
t
i
e
n
t
s,”
N
e
u
r
o
-
O
n
c
o
l
o
g
y
P
ra
c
t
i
c
e
,
v
o
l
.
7
,
n
o
.
5
,
p
p
.
4
9
8
–
5
0
6
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
9
3
/
n
o
p
/
n
p
a
a
0
1
5
.
[
4
]
P
.
Y
.
W
e
n
e
t
a
l
.
,
“
U
p
d
a
t
e
d
r
e
sp
o
n
se
a
ssessme
n
t
c
r
i
t
e
r
i
a
f
o
r
h
i
g
h
-
g
r
a
d
e
g
l
i
o
mas:
R
e
s
p
o
n
se
a
ssess
me
n
t
i
n
n
e
u
r
o
-
o
n
c
o
l
o
g
y
w
o
r
k
i
n
g
g
r
o
u
p
,
”
J
o
u
r
n
a
l
o
f
C
l
i
n
i
c
a
l
O
n
c
o
l
o
g
y
,
v
o
l
.
2
8
,
n
o
.
1
1
,
p
p
.
1
9
6
3
–
1
9
7
2
,
2
0
1
0
,
d
o
i
:
1
0
.
1
2
0
0
/
JC
O
.
2
0
0
9
.
2
6
.
3
5
4
1
.
[
5
]
E.
A
.
S
.
E
l
-
D
a
h
sh
a
n
,
H
.
M
.
M
o
h
s
e
n
,
K
.
R
e
v
e
t
t
,
a
n
d
A
.
B
.
M
.
S
a
l
e
m
,
“
C
o
mp
u
t
e
r
-
a
i
d
e
d
d
i
a
g
n
o
si
s
o
f
h
u
ma
n
b
r
a
i
n
t
u
mo
r
t
h
r
o
u
g
h
M
R
I
:
A
su
r
v
e
y
a
n
d
a
n
e
w
a
l
g
o
r
i
t
h
m,
”
E
x
p
e
rt
S
y
s
t
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
4
1
,
n
o
.
1
1
,
p
p
.
5
5
2
6
–
5
5
4
5
,
2
0
1
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
1
4
.
0
1
.
0
2
1
.
[
6
]
J.
K
a
n
g
,
Z.
U
l
l
a
h
,
a
n
d
J
.
G
w
a
k
,
“
M
r
i
-
b
a
s
e
d
b
r
a
i
n
t
u
m
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
e
n
s
e
mb
l
e
o
f
d
e
e
p
f
e
a
t
u
r
e
s
a
n
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
c
l
a
ss
i
f
i
e
r
s,”
S
e
n
s
o
rs
,
v
o
l
.
2
1
,
n
o
.
6
,
p
p
.
1
–
2
1
,
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
1
0
6
2
2
2
2
.
[
7
]
S
.
A
h
ma
d
a
n
d
P
.
K
.
C
h
o
u
d
h
u
r
y
,
“
O
n
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
d
e
e
p
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
n
e
t
w
o
r
k
s
f
o
r
b
r
a
i
n
t
u
mo
r
d
e
t
e
c
t
i
o
n
u
si
n
g
M
R
i
ma
g
e
s,
”
I
E
EE
A
c
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
5
9
0
9
9
–
5
9
1
1
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
1
7
9
3
7
6
.
[
8
]
S
.
S
h
a
n
t
h
i
,
S
.
S
a
r
a
d
h
a
,
J.
A
.
S
m
i
t
h
a
,
N
.
P
r
a
sa
t
h
,
a
n
d
H
.
A
n
a
n
d
a
k
u
mar,
“
A
n
e
f
f
i
c
i
e
n
t
a
u
t
o
m
a
t
i
c
b
r
a
i
n
t
u
m
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
o
p
t
i
m
i
z
e
d
h
y
b
r
i
d
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
N
e
t
w
o
rks
,
v
o
l
.
3
,
p
p
.
1
8
8
–
1
9
6
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
j
i
n
.
2
0
2
2
.
1
1
.
0
0
3
.
[
9
]
I
.
B
.
S
a
n
t
o
s
o
,
Y
.
A
d
r
i
a
n
t
o
,
A
.
S
e
n
su
s
i
a
t
i
,
D
.
W
u
l
a
n
d
a
r
i
,
a
n
d
I
.
P
u
r
n
a
m
a
,
“
Ep
i
l
e
p
t
i
c
EEG
S
i
g
n
a
l
C
l
a
s
si
f
i
c
a
t
i
o
n
U
si
n
g
C
o
n
v
o
l
u
t
i
o
n
a
l
N
e
u
r
a
l
N
e
t
w
o
r
k
B
a
s
e
d
o
n
M
u
l
t
i
-
S
e
g
m
e
n
t
o
f
E
EG
S
i
g
n
a
l
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
E
n
g
i
n
e
e
r
i
n
g
a
n
d
S
y
s
t
e
m
s
,
v
o
l
.
1
4
,
n
o
.
3
,
p
p
.
1
6
0
–
1
7
6
,
J
u
n
.
2
0
2
1
,
d
o
i
:
1
0
.
2
2
2
6
6
/
i
j
i
e
s
2
0
2
1
.
0
6
3
0
.
1
5
.
[
1
0
]
M
.
R
i
z
w
a
n
,
A
.
S
h
a
b
b
i
r
,
A
.
R
.
Ja
v
e
d
,
M
.
S
h
a
b
b
i
r
,
T.
B
a
k
e
r
,
a
n
d
D
.
A
l
-
Ju
m
e
i
l
y
O
b
e
,
“
B
r
a
i
n
t
u
m
o
r
a
n
d
g
l
i
o
m
a
g
r
a
d
e
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
G
a
u
ss
i
a
n
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
EEE
A
c
c
e
s
s
,
v
o
l
.
1
0
,
p
p
.
2
9
7
3
1
–
2
9
7
4
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
1
5
3
1
0
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.