I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
39
,
No
.
1
,
J
u
l
y
2
0
2
5
,
p
p
.
1
48
~
1
58
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/ijeecs.v
39
.i
1
.
p
p
1
48
-
1
58
148
J
o
ur
na
l
ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
Using
ResNe
t
a
rc
hitectu
re
with
M
RI for
cla
ss
ifica
ti
o
n of
bra
in ima
g
es
Su
bra
m
a
nia
n Dha
na
la
k
s
hm
i,
Su
bra
m
a
nia
n Ar
uls
elv
i
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
mm
u
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
B
h
a
r
a
t
h
I
n
st
i
t
u
t
e
o
f
H
i
g
h
e
r
E
d
u
c
a
t
i
o
n
a
n
d
R
e
s
e
a
r
c
h
,
C
h
e
n
n
a
i
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
19
,
2
0
2
4
R
ev
is
ed
Feb
15
,
2
0
2
5
Acc
ep
ted
Mar
25
,
2
0
2
5
A
stro
n
g
c
las
sifica
ti
o
n
m
o
d
e
l
t
h
a
t
c
a
n
c
o
rre
c
tl
y
d
e
tec
t
a
b
n
o
rm
a
li
ti
e
s
a
n
d
n
e
u
ro
l
o
g
ica
l
d
is
o
rd
e
rs
in
b
ra
i
n
i
m
a
g
e
s
is
th
e
m
a
in
g
o
a
l.
T
h
e
f
o
c
u
s
o
f
t
h
i
s
re
se
a
rc
h
is
o
n
imp
r
o
v
i
n
g
th
e
a
c
c
u
ra
c
y
o
f
M
RI
b
ra
in
ima
g
e
c
a
te
g
o
riza
ti
o
n
u
sin
g
re
sid
u
a
l
n
e
two
r
k
s
(
Re
sN
e
t
)
me
th
o
d
s.
Im
p
ro
v
i
n
g
t
h
e
m
o
d
e
l'
s
c
a
p
a
c
it
y
to
e
x
trac
t
c
o
m
p
lex
c
h
a
ra
c
teristi
c
s
fro
m
M
RI
ima
g
e
s
a
n
d
a
c
h
iev
in
g
m
o
re
a
c
c
u
ra
te
c
las
sifica
ti
o
n
re
su
lt
s
is
th
e
a
im
o
f
u
si
n
g
Re
sN
e
t
a
rc
h
it
e
c
tu
re
s.
B
y
c
o
n
d
u
c
ti
n
g
e
x
te
n
siv
e
e
x
p
e
rime
n
ts
a
n
d
v
a
li
d
a
ti
n
g
o
u
r
re
su
lt
s,
o
u
r
p
r
o
jec
t
a
ims
to
a
tt
a
in
to
p
-
n
o
tch
p
e
rfo
rm
a
n
c
e
in
b
ra
in
ima
g
e
c
las
sifica
ti
o
n
t
a
sk
s.
Th
e
g
o
a
l
is
t
o
h
e
l
p
im
p
ro
v
e
m
e
d
ica
l
d
iag
n
o
sis
a
n
d
trea
tme
n
t
p
la
n
n
i
n
g
.
A
se
c
o
n
d
a
ry
g
o
a
l
o
f
t
h
e
re
se
a
rc
h
is
to
d
e
term
in
e
if
d
e
e
p
lea
rn
in
g
a
p
p
r
o
a
c
h
e
s
h
a
v
e
a
n
y
u
se
in
ra
d
io
l
o
g
y
,
with
t
h
e
h
o
p
e
t
h
a
t
t
h
is
wil
l
lea
d
to
b
e
tt
e
r
m
e
d
ica
l
ima
g
e
a
n
a
ly
sis
p
ip
e
li
n
e
s.
T
h
e
m
a
in
o
b
jec
ti
v
e
is
to
m
a
k
e
it
e
a
sie
r
to
i
d
e
n
ti
f
y
n
e
u
ro
l
o
g
ica
l
p
r
o
b
lem
s
e
a
rly
o
n
,
wh
ich
wil
l
e
n
h
a
n
c
e
p
a
ti
e
n
t
o
u
tc
o
m
e
s
a
n
d
a
ll
o
w
f
o
r
m
o
re
c
a
lcu
late
d
trea
tme
n
t
d
e
c
isio
n
s.
Re
su
lt
s
p
ro
v
e
d
th
a
t
t
h
e
p
ro
p
o
se
d
Re
sN
e
t
sy
ste
m
a
c
h
iev
e
s
9
8
.
8
%
o
v
e
ra
ll
a
c
c
u
ra
c
y
with
9
8
.
6
%
se
n
siti
v
it
y
a
n
d
9
9
%
sp
e
c
ifi
c
it
y
.
K
ey
w
o
r
d
s
:
B
r
ain
I
m
ag
e
class
if
icatio
n
Ma
g
n
etic
r
eso
n
an
ce
im
a
g
in
g
Neu
r
o
lo
g
ical
d
is
o
r
d
er
s
R
esNet
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Su
b
r
am
an
ian
D
h
an
alak
s
h
m
i
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
B
h
ar
ath
I
n
s
titu
te
o
f
Hig
h
e
r
E
d
u
ca
tio
n
an
d
R
esear
ch
C
h
en
n
ai,
T
am
il Na
d
u
,
I
n
d
ia
E
m
ail: d
h
an
alak
s
h
m
is
n
r
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
o
a
d
d
r
e
s
s
t
h
e
c
o
m
p
l
e
x
p
r
o
b
le
m
o
f
e
f
f
i
ci
e
n
t
l
y
t
r
a
i
n
i
n
g
v
e
r
y
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
r
es
i
d
u
al
n
e
t
w
o
r
k
s
(
R
e
s
N
e
t
)
e
m
e
r
g
e
d
a
s
a
r
e
v
o
l
u
ti
o
n
a
r
y
n
e
u
r
a
l
n
e
t
w
o
r
k
d
e
s
i
g
n
[
1
]
.
T
h
e
p
i
o
n
e
e
r
i
n
g
w
o
r
k
o
f
K
a
i
m
i
n
g
H
e
e
t
a
l
.
i
n
2
0
1
5
l
a
i
d
t
h
e
g
r
o
u
n
d
w
o
r
k
f
o
r
R
e
s
Ne
t
,
a
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
o
p
t
i
m
i
z
at
i
o
n
a
n
d
d
e
s
i
g
n
f
r
a
m
e
w
o
r
k
t
h
a
t
i
n
t
r
o
d
u
c
e
d
a
n
e
w
n
o
t
i
o
n
c
a
l
le
d
r
e
s
i
d
u
a
l
le
a
r
n
i
n
g
.
T
h
e
d
e
g
r
a
d
a
t
i
o
n
p
r
o
b
l
e
m
w
a
s
a
m
a
j
o
r
r
o
a
d
b
l
o
c
k
f
o
r
t
h
e
t
r
a
d
i
t
i
o
n
a
l
d
e
e
p
n
e
tw
o
r
k
p
a
r
a
d
i
g
m
.
I
t
m
e
a
n
t
t
h
a
t
a
d
d
i
n
g
m
o
r
e
l
a
y
e
r
s
w
o
u
l
d
l
ea
d
t
o
w
o
r
s
e
p
e
r
f
o
r
m
a
n
c
e
b
e
c
a
u
s
e
o
f
t
h
i
n
g
s
l
i
k
e
v
a
n
is
h
i
n
g
g
r
a
d
i
en
t
s
a
n
d
o
p
t
i
m
i
za
t
i
o
n
c
o
m
p
l
e
x
ity
[
2
]
.
R
e
s
Ne
t
c
le
v
e
r
l
y
g
o
t
a
r
o
u
n
d
t
h
i
s
p
r
o
b
l
e
m
b
y
u
s
i
n
g
s
k
i
p
c
o
n
n
e
c
ti
o
n
s
,
s
o
m
et
im
e
s
c
a
ll
e
d
s
h
o
r
t
c
u
ts
.
T
h
es
e
l
et
t
h
e
n
e
t
w
o
r
k
s
k
i
p
o
v
e
r
s
p
e
ci
f
i
c
l
e
v
e
l
s
,
s
o
d
a
t
a
m
a
y
f
l
o
w
f
r
e
e
l
y
f
r
o
m
o
n
e
l
a
y
e
r
t
o
a
n
o
t
h
e
r
.
W
it
h
t
h
i
s
n
e
w
a
r
c
h
it
ec
t
u
r
a
l
d
e
s
i
g
n
,
R
es
N
et
v
a
r
i
a
n
ts
m
a
y
t
r
a
i
n
n
e
t
w
o
r
k
s
w
i
t
h
h
u
n
d
r
e
d
s
o
f
l
a
y
e
r
s
o
r
m
o
r
e
i
n
d
e
p
t
h
w
h
i
l
e
m
ai
n
t
a
i
n
i
n
g
o
r
i
m
p
r
o
v
i
n
g
p
e
r
f
o
r
m
a
n
c
e
m
e
t
r
i
c
s
[
3
]
.
T
o
p
r
e
v
e
n
t
g
r
a
d
i
e
n
t
s
f
r
o
m
d
is
a
p
p
e
a
r
i
n
g
d
u
r
i
n
g
t
r
a
i
n
i
n
g
,
R
e
s
N
et
i
n
c
o
r
p
o
r
a
t
es
r
e
s
i
d
u
a
l
co
n
n
e
c
t
i
o
n
s
,
w
h
i
c
h
e
f
f
e
c
t
i
v
el
y
s
o
l
v
es
t
h
e
v
a
n
i
s
h
i
n
g
g
r
a
d
i
e
n
t
i
s
s
u
e
[
4
]
.
T
h
i
s
d
e
s
i
g
n
e
l
e
m
e
n
t
i
m
p
r
o
v
e
s
t
h
e
n
e
t
w
o
r
k
'
s
g
e
n
e
r
a
l
i
z
a
b
i
li
t
y
b
y
r
e
d
u
c
i
n
g
o
p
t
i
m
i
z
a
ti
o
n
d
i
f
f
i
c
u
lt
i
es
a
n
d
i
n
c
r
e
a
s
i
n
g
c
o
n
v
e
r
g
e
n
c
e
r
a
t
es
v
ia
t
h
e
p
r
o
v
is
i
o
n
o
f
s
h
o
r
t
c
u
t
c
h
a
n
n
e
ls
f
o
r
g
r
a
d
ie
n
t
f
l
o
w
[
5
]
.
R
es
N
e
t
h
a
s
b
e
c
o
m
e
a
n
i
m
p
o
r
t
a
n
t
p
a
r
t
o
f
d
e
e
p
l
e
ar
n
i
n
g
a
n
d
h
a
s
b
e
e
n
u
s
e
d
e
x
t
e
n
s
i
v
e
l
y
i
n
m
a
n
y
d
i
f
f
e
r
e
n
t
f
i
e
l
d
s
,
s
u
c
h
a
s
s
e
m
a
n
t
ic
s
e
g
m
e
n
t
a
t
i
o
n
,
o
b
j
e
c
t
i
d
e
n
ti
f
i
c
a
t
i
o
n
,
a
n
d
p
i
c
t
u
r
e
c
l
a
s
s
i
f
ic
a
t
i
o
n
[
6
]
.
T
h
e
a
r
c
h
i
t
e
c
t
u
r
a
l
c
o
n
c
e
p
t
s
o
f
R
es
N
e
t
h
av
e
f
o
s
t
e
r
e
d
b
r
e
a
k
t
h
r
o
u
g
h
s
i
n
m
o
d
e
l
d
e
s
i
g
n
a
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Usi
n
g
R
esN
et
a
r
ch
itectu
r
e
w
it
h
MRI
fo
r
cla
s
s
ifica
tio
n
o
f b
r
a
in
ima
g
es
(
S
u
b
r
a
ma
n
ia
n
Dh
a
n
a
la
ksh
mi
)
149
o
p
t
i
m
i
z
a
ti
o
n
a
p
p
r
o
a
c
h
e
s
,
a
n
d
i
t
s
s
i
g
n
i
f
i
c
a
n
c
e
g
o
es
b
e
y
o
n
d
it
s
a
p
p
li
c
a
ti
o
n
s
i
n
c
o
m
p
u
t
e
r
v
is
i
o
n
t
o
o
t
h
e
r
f
i
e
l
d
s
w
i
t
h
i
n
a
r
t
i
f
i
c
i
al
i
n
t
e
l
li
g
e
n
c
e
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
[
7
]
.
S
e
c
t
i
o
n
3
d
e
l
v
e
s
i
n
t
o
t
h
e
M
R
I
f
o
r
C
l
a
s
s
i
f
i
c
at
i
o
n
o
f
B
r
a
i
n
I
m
a
g
e
s
a
n
d
h
o
w
R
es
N
et
m
o
d
e
ls
u
s
e
v
a
r
i
o
u
s
a
p
p
r
o
a
c
h
e
s
.
I
n
s
ec
t
i
o
n
4
,
a
n
e
x
a
m
p
l
e
o
f
a
M
R
I
f
o
r
C
l
as
s
i
f
i
c
a
ti
o
n
o
f
B
r
a
i
n
I
m
a
g
e
s
u
t
i
li
s
i
n
g
R
e
s
N
et
m
o
d
e
ls
wi
t
h
s
e
v
e
r
a
l
d
a
t
as
e
ts
a
r
e
u
t
il
i
s
e
d
.
Fi
n
a
l
l
y
,
i
n
s
e
cti
o
n
5
t
h
e
s
t
u
d
y
i
s
c
o
n
c
l
u
d
e
d
w
i
t
h
c
o
n
c
l
u
s
i
o
n
.
2.
L
I
T
E
RA
T
U
RE
SU
R
VE
Y
T
h
i
s
s
t
u
d
y
a
i
m
s
t
o
i
m
p
r
o
v
e
b
r
a
i
n
t
u
m
o
u
r
d
i
a
g
n
o
s
i
s
f
r
o
m
MR
I
i
m
a
g
e
s
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
e
s
p
e
ci
a
l
l
y
t
h
e
R
e
s
N
e
t
5
0
a
r
c
h
i
t
e
ct
u
r
e
a
n
d
g
r
a
d
i
e
n
t
-
w
e
i
g
h
t
e
d
c
l
a
s
s
a
c
t
iv
a
t
i
o
n
m
a
p
p
i
n
g
(
G
r
a
d
-
C
A
M
)
[
8
]
.
D
L
t
e
c
h
n
i
q
u
e
s
f
i
n
e
-
t
u
n
e
p
r
e
-
t
r
a
i
n
e
d
m
o
d
e
ls
l
i
k
e
R
e
s
N
e
t
,
a
n
a
d
v
a
n
c
e
d
c
la
s
s
i
f
i
c
at
i
o
n
m
e
t
h
o
d
.
R
es
N
e
t
-
b
a
s
e
d
t
r
a
n
s
f
o
r
m
e
r
s
e
s
t
i
m
a
t
e
b
r
a
i
n
a
g
e
a
n
d
c
l
ass
i
f
y
A
lz
h
e
i
m
e
r
'
s
d
is
e
as
e
.
C
N
N'
s
a
u
t
o
m
a
t
e
d
f
ea
t
u
r
e
e
x
tr
a
c
t
i
o
n
m
a
y
b
o
o
s
t
p
e
r
f
o
r
m
a
n
c
e
.
C
N
N
a
p
p
r
o
a
c
h
e
s
i
n
c
l
u
d
e
R
es
N
et
.
A
R
es
N
e
t
f
e
e
d
-
f
o
r
w
a
r
d
s
i
g
n
a
l
b
y
p
a
s
s
es
th
e
C
N
N
b
l
o
c
k
a
n
d
d
i
r
e
c
t
l
y
i
n
f
l
u
e
n
c
e
s
o
u
t
p
u
ts
.
R
e
s
N
et
r
e
p
l
ac
e
s
C
N
N
,
w
h
i
c
h
co
m
b
i
n
e
s
r
es
i
d
u
a
l
b
l
o
c
k
s
[
9
]
.
R
e
c
e
n
t
a
d
v
a
n
c
es
i
n
m
e
d
i
c
a
l
i
m
a
g
i
n
g
t
ec
h
n
o
l
o
g
y
h
a
v
e
m
a
d
e
MR
I
a
n
e
s
s
e
n
ti
a
l
t
o
o
l
f
o
r
c
l
i
n
ic
a
l
d
i
a
g
n
o
s
is
a
n
d
t
r
e
a
t
m
e
n
t
.
M
R
I
u
s
es
n
u
c
l
e
a
r
m
a
g
n
e
t
i
c
r
es
o
n
a
n
c
e
t
o
p
r
o
d
u
c
e
h
i
g
h
-
d
e
f
i
n
i
t
i
o
n
p
i
c
t
u
r
e
s
o
f
h
u
m
a
n
t
i
s
s
u
es
.
I
t
p
r
o
v
i
d
e
s
p
r
e
c
i
s
e
a
n
a
t
o
m
i
c
al
s
t
r
u
c
t
u
r
e
a
n
d
p
a
t
h
o
l
o
g
i
c
a
l
i
n
f
o
r
m
a
t
i
o
n
w
i
t
h
o
u
t
r
a
d
i
at
i
o
n
,
i
n
v
a
s
i
v
e
n
es
s
,
o
r
m
u
l
t
i
-
d
i
m
e
n
s
i
o
n
a
lit
y
.
W
i
t
h
m
o
r
e
MR
I
d
a
t
a
,
m
a
n
u
a
l
p
r
o
c
e
s
s
i
n
g
a
n
d
i
n
t
e
r
p
r
e
t
a
t
i
o
n
a
r
e
t
i
m
e
-
c
o
n
s
u
m
i
n
g
a
n
d
s
u
b
j
e
c
t
i
v
e
[
1
0
]
.
C
o
m
b
i
n
i
n
g
R
e
s
Ne
t
-
5
0
'
s
f
e
a
t
u
r
e
e
x
t
r
a
ct
i
o
n
a
n
d
c
la
s
s
i
f
i
ca
t
i
o
n
w
it
h
Y
OL
O
v
5
'
s
o
b
j
e
c
t
i
d
e
n
t
i
f
i
c
at
i
o
n
s
k
il
ls
y
i
el
d
s
a
c
o
m
p
l
e
t
e
m
o
d
e
l
.
T
h
es
e
s
y
n
e
r
g
i
s
ti
c
e
f
f
e
ct
s
h
e
l
p
e
d
t
h
e
m
o
d
e
l
a
c
c
u
r
a
te
l
y
i
d
e
n
ti
f
y
g
l
i
o
m
a
t
u
m
o
u
r
s
i
t
es
a
n
d
d
e
t
e
c
t
ti
n
y
d
e
t
a
ils
n
e
e
d
e
d
f
o
r
c
a
t
e
g
o
r
i
z
a
ti
o
n
.
YO
L
O
v
5
'
s
l
o
c
a
li
z
a
ti
o
n
w
as
i
m
p
r
o
v
e
d
u
s
i
n
g
R
es
N
et
-
5
0
'
s
f
e
at
u
r
e
e
x
t
r
a
c
ti
o
n
a
n
d
c
l
a
s
s
i
f
ic
a
t
i
o
n
[
1
1
]
.
A
R
e
s
N
e
t
5
0
-
i
m
p
r
o
v
e
d
b
r
a
i
n
t
u
m
o
u
r
c
l
a
s
s
i
f
ic
a
t
i
o
n
d
e
e
p
l
e
a
r
n
i
n
g
n
e
t
w
o
r
k
u
s
e
s
f
l
i
p
p
i
n
g
,
r
o
t
a
t
i
o
n
,
a
n
d
t
r
a
n
s
l
at
i
o
n
t
o
en
h
an
ce
d
ata.
M
R
I
b
r
ain
tu
m
o
u
r
s
eg
m
en
tatio
n
u
s
in
g
C
NN
-
en
h
an
ce
d
R
esNet5
0
an
d
U
-
Net
[
1
2
]
.
So
m
e
m
o
d
els we
r
e
VGG
-
1
6
,
R
esNet
-
5
0
,
an
d
E
f
f
icien
tNet
-
B
0
.
VGG
-
1
6
,
R
esNet
-
5
0
,
an
d
I
n
ce
p
tio
n
v
3
m
o
d
els
with
C
NN
p
r
e
-
tr
ain
in
g
wer
e
u
s
ed
to
au
to
m
atica
lly
p
r
ed
ict
an
d
lab
el
b
r
ain
tu
m
o
u
r
s
.
E
x
p
er
im
en
tal
r
esu
lts
s
h
o
w
th
at
R
esNet
-
5
0
o
u
tp
er
f
o
r
m
s
VGG
-
1
6
an
d
I
n
ce
p
tio
n
v
3
.
T
h
is
v
alid
ates
an
d
r
ec
o
m
m
en
d
s
R
esNet
-
5
0
f
o
r
tu
m
o
u
r
ca
teg
o
r
izatio
n
[
1
3
]
.
T
wo
-
ch
an
n
el
d
ee
p
n
e
u
r
al
n
et
wo
r
k
id
ea
f
o
r
tu
m
o
u
r
c
lass
if
icatio
n
th
at
is
m
o
r
e
ad
ap
tab
le
an
d
ef
f
ec
tiv
e.
I
n
itial
lo
ca
l
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
ar
e
ex
tr
ac
ted
u
s
in
g
I
n
ce
p
t
io
n
R
esNetV2
an
d
Xce
p
tio
n
n
etwo
r
k
s
'
co
n
v
o
lu
ti
o
n
b
l
o
ck
s
an
d
v
ec
to
r
ized
u
s
in
g
p
o
o
lin
g
-
b
ased
m
eth
o
d
s
[
1
4
]
.
T
r
an
s
f
er
lear
n
i
n
g
s
tu
d
y
u
s
ed
th
e
p
r
e
-
tr
ain
e
d
R
esNet5
0
ar
ch
itectu
r
e
to
ap
p
ly
co
n
tr
ast
s
tr
etch
in
g
an
d
h
is
to
g
r
am
eq
u
alis
atio
n
to
in
p
u
t
p
ictu
r
es
a
n
d
co
m
p
ar
e
ac
cu
r
ac
y
an
d
s
en
s
itiv
ity
.
R
esNet
d
if
f
er
s
f
r
o
m
VGG
an
d
Alex
Net.
R
esNet
's
m
icr
o
ar
ch
itectu
r
e
m
o
d
u
le
lay
o
u
t a
llo
ws ce
r
tain
lay
er
tr
an
s
iti
o
n
s
to
b
e
a
v
o
id
ed
an
d
o
th
er
s
t
o
b
e
m
a
d
e
[
1
5
]
.
Usi
n
g
a
C
NN
-
b
ased
n
etwo
r
k
to
d
etec
t
b
r
ain
ca
n
ce
r
in
M
R
I
s
was
s
u
g
g
ested
.
Den
s
e
E
f
f
icien
t
Net
o
u
tp
er
f
o
r
m
ed
R
esNet
-
5
0
,
Mo
b
ile
Net,
an
d
Mo
b
ileNetV2
.
R
esNet,
s
h
o
r
t
f
o
r
r
esid
u
al
n
etwo
r
k
,
s
o
lv
es
co
m
p
u
ter
v
is
io
n
d
if
f
ic
u
ltie
s
.
R
esNet1
0
1
's
3
3
b
lo
ck
s
o
f
1
0
4
co
n
v
o
lu
tio
n
al
lay
er
s
r
ec
y
cle
2
9
s
q
u
a
r
es
[
1
6
]
.
T
h
e
s
tu
d
y
ex
am
in
es
d
ee
p
co
n
v
o
lu
tio
n
al
lay
er
s
in
SR
C
NN
ar
ch
i
tectu
r
e
to
lear
n
co
m
p
lex
c
h
ar
ac
ter
is
tics
b
etwe
en
lo
w
-
an
d
h
ig
h
-
r
eso
lu
tio
n
p
h
o
t
o
p
atch
es.
Hier
ar
ch
ically
lear
n
in
g
n
etwo
r
k
s
lin
k
lo
w
-
r
eso
l
u
tio
n
p
atch
es
to
h
ig
h
-
r
eso
lu
tio
n
p
atch
es
with
o
u
t
in
ter
m
ed
iate
p
h
ases
.
B
r
ain
MRI
p
ictu
r
es
m
ay
b
e
cr
ea
ted
u
s
in
g
C
NNs,
Mo
b
ileNetV2
,
R
esNet1
5
2
V2
,
an
d
GAN
-
b
ased
au
g
m
en
tati
o
n
.
A
m
ix
e
d
co
n
v
o
lu
tio
n
s
m
eth
o
d
is
s
u
g
g
ested
[
1
7
]
.
Mic
r
o
s
o
f
t
R
esear
ch
cr
e
ated
th
e
1
5
2
-
lay
er
R
esNet
-
1
5
2
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
e
two
r
k
.
I
n
its
m
ai
n
in
n
o
v
atio
n
,
r
esid
u
al
co
n
n
ec
ti
o
n
s
o
r
s
k
ip
co
n
n
ec
tio
n
s
allo
w
th
e
n
etwo
r
k
to
lear
n
r
esid
u
al
f
u
n
ctio
n
s
,
m
ak
in
g
d
ee
p
n
etwo
r
k
tr
ai
n
in
g
s
im
p
l
er
.
I
m
a
g
e
ca
teg
o
r
izatio
n
a
n
d
o
b
ject
id
e
n
tific
atio
n
b
e
n
ef
it
f
r
o
m
R
esNet
-
1
5
2
'
s
d
ep
th
,
wh
ic
h
ex
tr
ac
ts
s
u
b
tle
f
ea
tu
r
es
an
d
p
atter
n
s
[
1
8
]
.
T
u
m
o
u
r
s
in
th
e
b
r
ain
m
ay
af
f
ec
t
b
r
ain
f
u
n
ctio
n
an
d
o
f
f
er
m
ajo
r
h
ea
lt
h
h
az
ar
d
s
.
T
i
m
ely
b
r
ain
tu
m
o
u
r
d
iag
n
o
s
is
i
s
es
s
en
tial
f
o
r
th
er
ap
y
.
B
r
ain
MRIs
ar
e
e
s
s
en
tial.
Do
cto
r
s
m
ay
s
ee
b
r
ai
n
ab
n
o
r
m
alities
v
ia
MRI
s
ca
n
s
,
wh
ich
u
s
e
s
tr
o
n
g
m
ag
n
ets an
d
r
ad
i
o
w
av
es [
1
9
]
.
T
o
p
r
o
d
u
ce
ac
cu
r
ate
an
d
tr
u
s
two
r
th
y
class
if
icatio
n
r
esu
lts
,
R
esNet3
4
,
R
esNet5
0
,
R
esNe
t1
0
1
,
an
d
R
esNet1
5
2
ar
e
em
p
lo
y
ed
.
co
m
p
licated
ar
ch
itectu
r
es
h
av
e
b
ee
n
u
s
ed
to
e
x
tr
ac
t
f
ea
tu
r
es
f
r
o
m
co
m
p
licated
p
ictu
r
es
in
g
en
er
ic
im
a
g
e
n
et
wo
r
k
s
lik
e
I
m
ag
eNe
t.
Ou
r
s
o
lu
t
io
n
o
v
e
r
co
m
es
th
e
f
a
d
in
g
is
s
u
e
in
d
ee
p
n
e
u
r
al
n
etwo
r
k
tr
ain
in
g
b
y
u
s
in
g
th
e
r
em
ain
in
g
R
esNet
d
esig
n
e
lem
en
ts
.
R
es
Net3
4
,
R
esNet5
0
,
R
esNet1
0
1
,
an
d
R
esNet1
5
2
h
av
e
3
4
,
5
0
,
1
0
1
,
an
d
1
5
2
lay
e
r
s
;
d
ep
th
an
d
r
ep
r
esen
tatio
n
ca
p
ac
ity
v
ar
y
[
2
0
]
.
Selectio
n
o
f
p
r
e
-
tr
ain
ed
m
o
d
els
lik
e
R
esn
et
i
s
d
if
f
icu
lt
o
win
g
to
m
an
y
f
a
cto
r
s
.
Sin
g
le
-
s
o
u
r
ce
f
ea
tu
r
es
ar
e
in
s
u
f
f
icien
t
to
ca
lcu
late
m
ax
im
al
ac
cu
r
ac
y
,
h
o
wev
er
th
is
m
eth
o
d
ad
d
e
d
r
ed
u
n
d
a
n
t
in
f
o
r
m
atio
n
a
n
d
p
r
o
ce
s
s
in
g
tim
e.
Fin
e
-
tu
n
in
g
th
e
R
esNet1
0
1
m
o
d
el
f
o
r
MRI
s
eq
u
en
ce
class
if
ic
ati
o
n
b
y
m
o
d
ality
is
d
escr
ib
ed
[
2
1
]
.
On
ly
m
ag
n
etic
r
eso
n
an
ce
im
a
g
in
g
(
MRI)
ca
n
ass
ess
ce
ll
an
d
tis
s
u
e
b
i
o
c
h
em
ical
an
d
m
etab
o
lic
s
tatu
s
with
o
u
t
h
ar
m
b
y
ex
am
in
in
g
tis
s
u
e
s
tr
u
ctu
r
e.
First,
it
m
ay
m
ea
s
u
r
e
tu
m
o
u
r
wo
r
s
en
in
g
,
t
h
en
in
v
asio
n
,
an
d
f
in
ally
b
r
ain
t
u
m
o
u
r
s
tate
[
2
2
]
.
T
h
e
R
esNet
-
5
0
d
esig
n
is
u
p
d
ated
to
ex
tr
ac
t
c
r
itical
in
f
o
r
m
atio
n
u
s
in
g
a
s
elf
-
atten
tio
n
b
lo
c
k
.
B
ay
esian
o
p
tim
is
atio
n
(
B
O)
o
p
tim
is
ed
th
e
h
y
p
e
r
p
ar
am
ete
r
s
,
wh
ich
wer
e
u
s
ed
to
tr
ain
th
e
m
o
d
el
a
n
d
ex
tr
ac
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
1
48
-
1
58
150
f
ea
tu
r
es.
A
f
in
e
-
t
u
n
ed
R
esNet
-
5
0
ar
c
h
itect
u
r
e
with
a
s
elf
-
atten
tio
n
lay
er
was
tr
ain
ed
f
r
o
m
s
cr
atch
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
[
2
3
]
.
R
esid
u
al
n
eu
r
al
n
etwo
r
k
s
b
u
il
d
n
etwo
r
k
s
f
r
o
m
r
esid
u
al
b
lo
c
k
s
.
A
5
0
-
lay
er
C
NN
is
R
esNe
t
-
5
0
.
T
h
e
50
-
lay
er
C
NN
h
as
4
8
co
n
v
o
lu
tio
n
al
lay
er
s
,
1
Ma
x
Po
o
l,
an
d
1
av
e
r
ag
e
p
o
o
l.
A
50
-
la
y
er
C
NN
h
as
o
n
e
Ma
x
Po
o
l
lay
er
,
o
n
e
av
e
r
ag
e
p
o
o
l
lay
e
r
,
an
d
4
8
c
o
n
v
o
lu
tio
n
al
lay
er
s
.
R
esNet
is
a
n
eu
r
al
n
etwo
r
k
th
at
u
n
d
er
p
in
s
s
ev
er
al
co
m
p
u
ter
v
is
io
n
ap
p
licatio
n
s
[
2
4
]
.
R
esNet5
0
u
s
es
r
esid
u
al
co
n
n
ec
ti
o
n
s
in
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
.
R
esNet
-
5
0
h
as
4
8
co
n
v
o
lu
tio
n
al
lay
e
r
s
,
1
Ma
x
Po
o
l
lay
er
,
an
d
1
a
v
er
ag
e
p
o
o
l
lay
e
r
.
R
esNet
iter
atio
n
s
s
h
ar
e
a
co
n
ce
p
t
b
u
t h
av
e
d
if
f
e
r
en
t
lay
er
co
u
n
ts
.
R
esn
et5
0
ca
n
h
an
d
le
5
0
n
e
u
r
al
n
etwo
r
k
lay
er
s
[
2
5
]
.
A
th
o
r
o
u
g
h
co
m
p
ar
is
o
n
o
f
tr
an
s
f
er
lear
n
i
n
g
-
b
ased
C
NN
m
o
d
els
p
r
e
-
tr
ain
ed
u
s
in
g
VGG1
6
,
R
esNet
-
5
0
,
an
d
I
n
ce
p
tio
n
V3
ar
ch
itectu
r
es
f
o
r
b
r
ain
tu
m
o
u
r
ce
ll
p
r
ed
ictio
n
.
I
t
u
s
ed
th
r
esh
o
ld
i
n
g
a
n
d
wate
r
s
h
ed
s
eg
m
en
tatio
n
[
2
6
]
.
C
NN
-
b
ased
Mo
d
if
ied
R
esNet1
5
2
v
2
class
if
ies
b
r
ain
s
t
r
o
k
e
C
T
im
ag
es
a
p
p
r
o
p
r
iately
.
T
h
is
s
tu
d
y
g
iv
es
au
to
m
ated
s
tr
o
k
e
d
iag
n
o
s
tic
an
d
p
r
e
v
en
tiv
e
m
et
h
o
d
s
f
o
r
in
d
iv
id
u
al
h
ea
lth
a
n
d
well
-
b
ein
g
[
2
7
]
.
B
r
ain
tu
m
o
u
r
class
if
icatio
n
u
s
in
g
R
esNet
(
2
+1
)
D
an
d
R
esNet
Mix
ed
C
o
n
v
o
lu
tio
n
.
T
h
ey
i
n
tr
o
d
u
ce
d
R
esNet
(
2
+1
)
D
an
d
R
esNet
Mix
ed
C
o
n
v
o
lu
tio
n
,
wh
ich
u
s
ed
2
D
an
d
3
D
c
o
n
v
o
lu
tio
n
.
T
h
e
two
m
o
d
els
o
u
tp
er
f
o
r
m
ed
R
esNet3
D
in
th
eir
test
in
g
[
2
8
]
.
I
n
b
r
ain
tu
m
o
u
r
d
etec
tio
n
,
Alex
Net,
Go
o
g
leNe
t,
an
d
R
esNet
-
18
wer
e
co
m
p
ar
ed
.
T
h
is
is
d
o
n
e
b
y
co
m
p
a
r
in
g
f
o
u
r
Ker
as
m
o
d
els:
R
e
s
Net5
0
,
Den
s
eNe
t2
0
1
,
I
n
ce
p
tio
n
V3
,
a
n
d
Mo
b
ileNet.
T
h
e
co
m
p
ar
is
o
n
d
eter
m
in
es
th
e
b
e
s
t
d
ee
p
lear
n
in
g
m
o
d
el
f
o
r
th
e
jo
b
[
2
9
]
.
W
o
v
en
clo
th
p
atter
n
id
en
tific
atio
n
u
s
in
g
R
esNet
-
5
0
.
R
esNet
o
u
tp
er
f
o
r
m
s
o
th
er
ap
p
r
o
ac
h
es.
I
n
tr
an
s
f
er
lear
n
in
g
cl
o
th
p
atter
n
r
ec
o
g
n
itio
n
,
o
v
er
f
itti
n
g
is
co
m
m
o
n
.
Usi
n
g
th
e
b
ac
k
d
r
o
p
a
b
o
v
e,
th
is
wo
r
k
p
r
o
p
o
s
es
a
R
esNet
m
o
d
el
with
d
r
o
p
o
u
t
r
eg
u
lar
is
atio
n
an
d
ex
am
in
es
its
im
p
ac
t
o
n
Palem
b
a
n
g
s
o
n
g
k
et
f
ab
r
ic
m
o
tif
p
ictu
r
e
r
ec
o
g
n
itio
n
with
d
ata
au
g
m
en
tatio
n
[
3
0
]
.
T
r
an
s
f
er
lear
n
in
g
,
u
s
in
g
R
esNet
an
d
L
eNe
t
m
o
d
el
to
p
o
lo
g
ies,
is
th
e
n
ex
t
s
tag
e.
T
r
an
s
f
er
lea
r
n
in
g
h
el
p
s
u
s
to
a
p
p
ly
f
ea
tu
r
es
ac
q
u
ir
e
d
o
n
lar
g
e
d
atasets
to
o
u
r
b
atik
class
if
icatio
n
ch
allen
g
e
b
y
u
s
in
g
ar
ch
itectu
r
al
k
n
o
wled
g
e
[
3
1
]
.
Pro
p
o
s
ed
iOS
p
r
o
t
o
ty
p
e
u
s
es
p
ictu
r
e
co
m
p
ar
is
o
n
a
n
d
tex
t
m
atch
in
g
.
As
d
is
cu
s
s
ed
later
in
th
e
p
ap
er
,
it
u
s
es
th
e
R
esNe
t
-
5
0
ar
ch
itectu
r
e
f
o
r
im
ag
e
f
ea
tu
r
e
ex
tr
a
ctio
n
o
win
g
to
its
b
en
ef
its
an
d
g
r
ea
ter
p
er
f
o
r
m
an
ce
th
an
a
p
r
io
r
d
esig
n
.
Pictu
r
e
s
im
ilar
ity
is
ca
lcu
lated
u
s
in
g
E
u
clid
ea
n
d
is
tan
ce
.
L
o
s
t
item
r
ep
o
r
ts
u
s
e
co
s
in
e
s
im
ilar
ity
an
d
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
p
r
e
-
p
r
o
ce
s
s
tex
t
in
s
tr
in
g
m
atch
in
g
[
3
2
].
3.
M
E
T
H
O
D
Patter
n
r
ec
o
g
n
itio
n
is
o
f
ten
d
e
s
cr
ib
ed
as th
e
s
tu
d
y
o
f
m
ea
s
u
r
in
g
th
in
g
s
in
m
ac
h
in
e
lear
n
i
n
g
.
T
h
e
g
o
al
o
f
th
is
s
cien
tific
d
is
cip
lin
e
is
to
d
ev
elo
p
d
ec
is
io
n
-
m
a
k
in
g
t
o
o
ls
u
s
in
g
a
s
et
o
f
p
r
e
v
io
u
s
ly
e
s
tab
lis
h
ed
m
etr
ics,
o
f
ten
r
ef
er
r
e
d
to
as
tr
ain
i
n
g
d
a
ta.
Her
e,
t
h
e
test
d
ata
is
s
o
r
ted
in
to
o
n
e
o
f
s
ev
er
al
p
r
e
-
estab
li
s
h
ed
class
es
u
s
in
g
th
e
in
f
er
r
e
d
ju
d
g
m
en
t te
ch
n
iq
u
e.
3
.
1
.
ResNet
1
8
I
n
(
1
)
s
h
o
ws
th
e
R
esn
et1
8
,
w
h
er
e
in
d
icate
s
th
e
n
etwo
r
k
'
s
lear
n
t
r
esid
u
al
f
u
n
ctio
n
,
wh
ich
is
u
s
u
ally
m
ad
e
u
p
o
f
c
o
n
v
o
lu
t
io
n
al
la
y
er
s
,
b
atch
n
o
r
m
aliza
tio
n
,
an
d
R
eL
U
ac
tiv
atio
n
s
.
T
h
e
r
esi
d
u
al
co
n
n
ec
tio
n
is
f
o
r
m
ed
b
y
ad
d
in
g
th
e
in
p
u
t
t
o
th
e
o
u
tp
u
t;
th
is
allo
ws
th
e
g
r
ad
ien
ts
to
f
lo
w
m
o
r
e
ea
s
ily
d
u
r
i
n
g
tr
ain
in
g
.
B
y
r
ed
u
cin
g
th
e
im
p
ac
t
o
f
th
e
v
an
is
h
in
g
g
r
a
d
ien
t
is
s
u
e,
th
is
eq
u
ati
o
n
ca
p
tu
r
es
R
esNet1
8
'
s
co
r
e
co
n
ce
p
t
an
d
m
ak
es o
p
tim
izatio
n
o
f
d
ee
p
er
n
etwo
r
k
s
s
im
p
ler
.
=
(
)
+
(
1
)
W
ith
in
th
e
r
ea
lm
o
f
co
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
d
esig
n
,
R
esNet
1
8
,
wh
ich
is
an
ab
b
r
ev
iatio
n
f
o
r
R
esid
u
al
Netwo
r
k
with
1
8
lay
er
s
,
is
g
en
er
ally
u
s
ed
f
o
r
th
e
p
u
r
p
o
s
e
o
f
im
ag
e
class
if
icatio
n
task
s
.
T
h
e
d
if
f
icu
lty
o
f
t
r
ain
in
g
ex
tr
em
el
y
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
is
a
d
d
r
ess
ed
b
y
t
h
is
s
o
lu
tio
n
,
wh
ich
was
d
ev
elo
p
ed
b
y
Kaim
in
g
He
an
d
c
o
lleag
u
es.
I
t
d
o
es
th
is
b
y
i
n
co
r
p
o
r
atin
g
r
esid
u
al
co
n
n
ec
tio
n
s
.
T
h
ese
c
o
n
n
ec
tio
n
s
m
a
k
e
it
p
o
s
s
ib
le
f
o
r
g
r
ad
ien
ts
to
f
lo
w
s
tr
aig
h
t
ac
r
o
s
s
th
e
lay
er
s
,
wh
ich
h
elp
to
allev
iate
th
e
is
s
u
e
o
f
d
is
a
p
p
ea
r
in
g
g
r
ad
ien
ts
an
d
m
a
k
e
it
p
o
s
s
ib
le
to
tr
ain
d
ee
p
er
n
etwo
r
k
s
in
a
m
o
r
e
ef
f
icien
t
m
a
n
n
er
.
T
h
e
ar
ch
itectu
r
e
o
f
R
esNet
1
8
is
m
ad
e
u
p
o
f
a
s
tack
o
f
co
n
v
o
lu
tio
n
al
lay
er
s
,
wh
ich
is
th
en
f
o
llo
wed
b
y
a
n
u
m
b
er
o
f
r
esid
u
al
b
lo
ck
s
f
o
r
f
u
r
th
er
p
r
o
ce
s
s
in
g
.
T
h
e
o
r
ig
in
al
in
p
u
t
is
ad
d
ed
to
th
e
o
u
tp
u
t
o
f
ea
c
h
r
esid
u
al
b
lo
ck
,
wh
ich
is
ac
co
m
p
lis
h
ed
b
y
t
h
e
u
s
e
o
f
s
h
o
r
tcu
t
co
n
n
ec
tio
n
s
,
also
k
n
o
wn
as
s
k
ip
co
n
n
ec
tio
n
s
.
E
a
ch
r
esid
u
al
b
lo
ck
co
n
s
is
ts
o
f
two
o
r
th
r
ee
co
n
v
o
lu
tio
n
al
lay
er
s
.
T
h
e
lear
n
in
g
o
f
r
esid
u
al
f
u
n
ctio
n
s
is
f
ac
ilit
ated
as
a
r
esu
lt
o
f
th
is
,
wh
ich
m
ak
es it sim
p
ler
f
o
r
th
e
n
etwo
r
k
to
lear
n
th
e
i
d
e
n
tity
m
ap
p
in
g
.
I
n
co
m
p
ar
is
o
n
to
o
th
er
v
ar
iat
io
n
s
,
s
u
ch
as
R
esNet
5
0
o
r
R
esNet
1
0
1
,
R
esNet
1
8
h
as
a
r
elativ
ely
s
h
allo
w
d
ep
th
,
wh
ich
allo
ws
it
to
estab
lis
h
a
co
m
p
r
o
m
is
e
b
et
wee
n
t
h
e
co
m
p
lex
ity
o
f
th
e
m
o
d
el
a
n
d
th
e
ass
o
ciate
d
co
m
p
u
tin
g
co
s
t.
I
t
h
as
n
o
t
o
n
ly
ac
h
iev
ed
s
tate
-
of
-
th
e
-
ar
t
r
esu
lts
in
im
ag
e
class
i
f
icatio
n
task
s
,
b
u
t
it
h
as
also
ex
h
i
b
ited
b
etter
p
e
r
f
o
r
m
an
ce
o
n
a
v
a
r
iety
o
f
b
en
ch
m
ar
k
d
atasets
,
in
clu
d
in
g
I
m
a
g
eNe
t.
Fu
r
th
e
r
m
o
r
e,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Usi
n
g
R
esN
et
a
r
ch
itectu
r
e
w
it
h
MRI
fo
r
cla
s
s
ifica
tio
n
o
f b
r
a
in
ima
g
es
(
S
u
b
r
a
ma
n
ia
n
Dh
a
n
a
la
ksh
mi
)
151
d
u
e
to
th
e
f
ac
t th
at
its
d
es
ig
n
is
b
o
th
s
im
p
le
an
d
ef
f
icien
t,
it
h
as b
ec
o
m
e
a
well
-
lik
ed
o
p
tio
n
f
o
r
th
e
p
u
r
p
o
s
e
o
f
tr
an
s
f
er
lear
n
in
g
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
in
c
o
m
p
u
te
r
v
is
io
n
ap
p
licatio
n
s
.
Her
e
t
h
e
Fig
u
r
e
1
p
r
esen
ts
th
e
R
esNet
1
8
ar
ch
itectu
r
e.
I
n
th
is
th
e
co
n
v
o
lu
tio
n
o
f
7
x
7
o
f
6
4
is
r
ep
ea
t
ed
o
n
ce
,
an
d
af
t
er
m
ax
p
o
o
lin
g
th
e
co
n
v
o
l
u
tio
n
o
f
3
x
3
o
f
6
4
is
r
ep
ea
ted
4
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
1
2
8
is
r
ep
ea
t
ed
4
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
3
x
3
o
f
2
5
6
is
r
ep
ea
ted
4
t
im
es,
th
e
c
o
n
v
o
lu
ti
o
n
o
f
3
x
3
o
f
5
1
2
i
s
r
ep
ea
ted
4
tim
es
.
Fig
u
r
e
1
.
R
esNet
1
8
ar
ch
itectu
r
e
3
.
2
.
ResNet
3
4
W
h
en
it
co
m
es
to
im
ag
e
cla
s
s
if
icatio
n
task
s
,
th
e
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
d
esig
n
k
n
o
wn
as
R
esNet
3
4
,
wh
ich
is
an
ab
b
r
ev
iatio
n
f
o
r
R
esid
u
al
Netwo
r
k
with
3
4
lay
er
s
,
is
we
ll
r
ec
o
g
n
ized
f
o
r
its
ef
f
icac
y
.
R
esNet
3
4
is
a
v
ar
iatio
n
o
f
th
e
R
esNet
f
am
ily
th
at
w
as
in
tr
o
d
u
ce
d
in
2
0
1
6
b
y
Kaim
in
g
He
an
d
co
lleag
u
es.
I
ts
p
u
r
p
o
s
e
is
to
s
o
lv
e
th
e
v
a
n
is
h
in
g
g
r
ad
ien
t
is
s
u
e
th
at
is
ex
p
er
ien
ce
d
in
d
ee
p
n
eu
r
al
n
etwo
r
k
s
.
Sk
ip
co
n
n
ec
tio
n
s
,
also
k
n
o
wn
as
s
h
o
r
tcu
ts
,
ar
e
in
clu
d
e
d
in
to
th
e
d
esig
n
.
T
h
ese
s
h
o
r
tcu
ts
en
ab
l
e
g
r
ad
ien
ts
to
f
lo
w
th
r
o
u
g
h
th
e
n
etwo
r
k
in
a
m
o
r
e
d
ir
ec
t
m
an
n
er
d
u
r
in
g
tr
a
in
in
g
,
wh
ich
h
elp
s
to
m
itig
ate
th
e
d
eg
r
ad
atio
n
p
r
o
b
lem
.
T
h
er
e
ar
e
a
to
tal
o
f
3
4
lay
er
s
th
at
m
ak
e
u
p
R
esNet
3
4
.
R
e
s
Net
3
4
h
as
h
ig
h
er
p
er
f
o
r
m
an
ce
in
co
m
p
ar
is
o
n
t
o
o
ld
e
r
co
n
v
o
lu
t
io
n
al
n
eu
r
al
n
etwo
r
k
d
esig
n
s
.
T
h
is
is
m
o
s
tly
b
ec
au
s
e
to
t
h
e
d
ep
th
an
d
s
k
ip
co
n
n
ec
tio
n
s
th
at
it
p
o
s
s
ess
es.
I
n
(
2
)
s
h
o
ws
th
e
R
esNet
3
4
,
wh
er
e
1
(
)
an
d
2
(
1
(
)
)
r
ep
r
esen
t
th
e
f
u
n
ctio
n
s
th
at
th
e
n
etwo
r
k
lear
n
t a
s
r
esi
d
u
als.
1
An
d
2
co
m
p
r
is
ed
o
f
m
a
n
y
co
n
v
o
l
u
tio
n
al
lay
er
s
th
at
ar
e
ac
tiv
ated
u
s
in
g
R
eL
U
an
d
n
o
r
m
alize
d
u
s
in
g
b
atch
in
g
.
T
h
e
in
p
u
t
x
is
u
s
ed
to
c
r
ea
te
th
e
r
esid
u
al
co
n
n
ec
tio
n
b
y
ad
d
in
g
it
b
ac
k
to
th
e
s
ec
o
n
d
r
esid
u
al
f
u
n
ct
io
n
'
s
o
u
tp
u
t.
W
ith
th
e
h
elp
o
f
th
is
eq
u
atio
n
,
R
esNet3
4
was
ab
le
to
lear
n
c
o
m
p
licated
m
ap
p
in
g
s
an
d
p
r
o
p
a
g
ate
g
r
a
d
ien
ts
ef
f
icien
t
ly
,
p
av
in
g
th
e
way
f
o
r
th
e
tr
ain
in
g
o
f
d
ee
p
er
ar
ch
itectu
r
es.
Fig
u
r
e
2
p
r
esen
ts
th
e
R
esNet3
4
b
lo
ck
d
ia
g
r
a
m
ar
ch
itectu
r
e.
I
n
th
is
th
e
co
n
v
o
lu
tio
n
o
f
7
x
7
o
f
6
4
is
r
ep
ea
ted
o
n
ce
,
an
d
a
f
ter
m
ax
p
o
o
lin
g
th
e
c
o
n
v
o
lu
tio
n
o
f
3
x
3
o
f
6
4
is
r
e
p
ea
t
ed
3
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
3
x
3
o
f
1
2
8
is
r
ep
ea
ted
4
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
2
5
6
is
r
ep
ea
t
e
d
6
tim
es,
th
e
c
o
n
v
o
lu
ti
o
n
o
f
3
x
3
o
f
5
1
2
is
r
ep
ea
te
d
3
tim
es.
=
2
(
1
(
)
)
+
(
2
)
Fig
u
r
e
2
.
R
esNet
34
ar
ch
itectu
r
e
3
.
3
.
ResNet
5
0
I
t
is
a
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
ar
c
h
itectu
r
e
th
at
i
s
f
r
eq
u
en
tly
u
tili
ze
d
in
co
m
p
u
ter
v
is
io
n
ap
p
licatio
n
s
,
n
o
tab
ly
i
n
im
a
g
e
class
if
icatio
n
an
d
o
b
ject
r
ec
o
g
n
itio
n
.
R
esNet
5
0
is
a
n
ab
b
r
ev
iatio
n
f
o
r
R
esid
u
al
Netwo
r
k
with
5
0
la
y
er
s
.
T
h
r
o
u
g
h
th
is
,
it
is
p
o
s
s
ib
le
to
tr
ain
n
etwo
r
k
s
th
at
a
r
e
f
ar
m
o
r
e
co
m
p
le
x
,
r
ea
ch
in
g
u
p
to
f
if
ty
lay
er
s
,
w
h
ile
y
et
p
r
eser
v
in
g
a
lev
el
o
f
co
m
p
u
tatio
n
al
co
m
p
lex
ity
th
at
is
to
ler
ab
le.
T
h
e
p
er
f
o
r
m
an
ce
o
f
R
esNet
5
0
h
a
s
b
ee
n
e
x
ce
p
tio
n
al
ac
r
o
s
s
a
v
a
r
iety
o
f
b
e
n
ch
m
ar
k
d
atasets
,
ex
ce
ed
in
g
th
at
o
f
ea
r
lier
m
o
d
els
th
at
we
r
e
c
o
n
s
id
er
ed
to
b
e
s
tate
-
of
-
th
e
-
ar
t.
B
ec
au
s
e
o
f
its
d
ep
t
h
a
n
d
ar
ch
ite
ctu
r
al
d
esig
n
,
it
h
as
b
ec
o
m
e
an
ess
en
tial
co
m
p
o
n
e
n
t
in
co
n
tem
p
o
r
ar
y
d
ee
p
lear
n
in
g
r
esear
ch
an
d
a
p
p
licatio
n
s
.
B
o
th
its
ef
f
icien
cy
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
1
48
-
1
58
152
an
d
its
ad
ap
tab
ilit
y
in
r
eso
lv
in
g
d
if
f
icu
lt
v
is
u
al
id
en
tific
atio
n
p
r
o
b
lem
s
ar
e
s
h
o
w
n
b
y
t
h
e
wid
esp
r
ea
d
u
s
ag
e
o
f
th
is
tech
n
o
lo
g
y
.
I
n
(
3
)
s
h
o
w
s
th
e
R
e
s
Net
5
0
wh
er
e
1
(
)
,
2
(
1
(
)
)
,
3
2
(
1
(
)
)
d
is
p
lay
th
e
n
etwo
r
k
'
s
r
em
ain
in
g
f
u
n
ctio
n
ali
ties
.
T
h
ese
o
p
er
atio
n
s
ar
e
b
u
ilt
u
s
in
g
lay
er
ed
c
o
n
v
o
lu
tio
n
al
lay
e
r
s
th
at
h
av
e
b
ee
n
ac
tiv
ated
u
s
in
g
R
eL
U
an
d
b
atch
n
o
r
m
aliza
tio
n
.
T
h
e
in
p
u
t
x
co
n
s
titu
tes
th
e
r
esid
u
al
co
n
n
ec
tio
n
b
y
b
ein
g
r
ea
p
p
lied
to
th
e
th
ir
d
r
esid
u
al
f
u
n
ctio
n
'
s
o
u
tp
u
t.
T
h
e
ar
ch
itectu
r
e
o
f
R
esNet5
0
is
en
ca
p
s
u
lated
in
th
is
eq
u
atio
n
,
wh
ich
allo
ws
it
to
lear
n
co
m
p
lex
f
ea
tu
r
e
r
ep
r
e
s
en
tatio
n
s
an
d
s
o
lv
e
th
e
d
is
ap
p
ea
r
in
g
g
r
ad
ien
t
p
r
o
b
lem
,
m
ak
in
g
d
ee
p
n
eu
r
al
n
etwo
r
k
tr
ai
n
in
g
m
u
ch
ea
s
ie
r
.
T
h
e
R
esNet
5
0
b
lo
ck
d
iag
r
am
ar
ch
itectu
r
e
is
p
r
esen
ted
in
Fig
u
r
e
3
.
I
n
th
is
th
e
co
n
v
o
l
u
tio
n
o
f
7
x
7
o
f
6
4
is
r
e
p
ea
ted
o
n
ce
,
an
d
af
te
r
m
ax
p
o
o
lin
g
,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
6
4
is
r
ep
ea
ted
3
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
6
4
is
r
ep
ea
ted
3
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
1
2
8
i
s
r
ep
ea
ted
4
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
1
2
8
is
r
ep
ea
ted
4
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
2
5
6
i
s
r
ep
ea
ted
6
tim
es
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
2
5
6
is
r
ep
ea
ted
6
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
5
1
2
i
s
r
ep
ea
ted
7
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
5
1
2
is
r
ep
ea
ted
3
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
1
0
2
4
is
r
ep
ea
ted
6
tim
es,
th
e
co
n
v
o
lu
t
io
n
o
f
1
x
1
o
f
2
0
4
8
is
r
ep
ea
ted
3
tim
es.
=
3
(
2
(
1
(
)
)
)
+
(
3
)
Fig
u
r
e
3
.
R
esNet5
0
ar
ch
itectu
r
e
3
.
4
.
ResNet
1
0
1
L
ik
e
R
esNet
-
5
0
b
u
t
1
0
1
lay
er
s
.
I
n
cr
ea
s
in
g
n
etwo
r
k
d
ep
t
h
im
p
r
o
v
es
p
er
f
o
r
m
an
ce
.
W
ith
its
1
0
1
lay
er
s
,
R
esNet
-
1
0
1
is
a
co
m
p
l
ex
R
esNet
m
o
d
el
th
at
ad
v
an
ce
s
d
ee
p
lear
n
in
g
.
R
esNet
-
1
0
1
s
o
lv
es th
e
v
an
is
h
in
g
g
r
ad
ien
t
p
r
o
b
lem
b
y
m
er
g
in
g
r
esid
u
al
b
lo
ck
s
with
id
en
tity
s
h
o
r
tcu
ts
,
allo
win
g
d
ee
p
n
etwo
r
k
tr
ain
in
g
.
C
o
n
v
o
lu
tio
n
al
la
y
er
s
,
b
atch
n
o
r
m
alis
atio
n
,
an
d
R
eL
U
ac
tiv
atio
n
s
in
ea
ch
r
esid
u
al
b
lo
c
k
im
p
r
o
v
e
tr
ain
i
n
g
s
tab
ilit
y
an
d
ef
f
icien
c
y
.
T
h
e
a
r
ch
itectu
r
e'
s
d
ep
th
an
d
b
o
ttlen
ec
k
lay
er
s
d
r
am
atica
lly
r
ed
u
c
e
p
ar
am
eter
s
wh
ile
r
etain
in
g
p
er
f
o
r
m
an
ce
.
R
esNet
-
1
0
1
is
s
u
ited
f
o
r
c
h
allen
g
in
g
ap
p
licatio
n
s
lik
e
h
i
g
h
-
r
eso
lu
tio
n
im
ag
e
class
if
icatio
n
,
o
b
ject
id
en
tific
atio
n
,
an
d
s
em
an
tic
s
eg
m
en
tat
io
n
b
ec
au
s
e
it
ex
tr
ac
ts
n
u
an
ce
d
f
ea
tu
r
es.
I
ts
g
r
ea
t
ac
cu
r
ac
y
an
d
r
esil
ien
ce
h
av
e
m
ad
e
it
a
v
ita
l
to
o
l
in
co
m
p
u
t
er
v
is
io
n
an
d
d
ee
p
lear
n
in
g
r
e
s
ea
r
ch
an
d
p
r
ac
tice.
T
h
e
Fig
u
r
e
4
s
h
o
ws
th
e
R
esNet
1
0
1
Ar
c
h
itectu
r
e.
I
n
th
is
th
e
co
n
v
o
lu
tio
n
o
f
7
x
7
o
f
6
4
is
r
ep
ea
ted
o
n
ce
,
an
d
af
ter
m
ax
p
o
o
lin
g
,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
6
4
is
r
ep
ea
ted
3
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
6
4
is
r
ep
ea
ted
3
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
1
x
1
o
f
128
is
r
e
p
ea
ted
4
tim
es,
t
h
e
co
n
v
o
lu
t
io
n
o
f
3
x
3
o
f
1
2
8
is
r
ep
ea
ted
4
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
1
x
1
o
f
2
5
6
is
r
ep
ea
ted
2
6
tim
es
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
2
5
6
is
r
e
p
ea
ted
23
tim
es,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
5
1
2
i
s
r
ep
ea
ted
7
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
3
x
3
o
f
5
1
2
is
r
ep
ea
ted
3
tim
es
,
th
e
co
n
v
o
l
u
tio
n
o
f
1
x
1
o
f
1
0
2
4
is
r
ep
ea
ted
2
3
tim
es,
th
e
co
n
v
o
lu
tio
n
o
f
1
x
1
o
f
2
0
4
8
is
r
ep
ea
ted
3
tim
es.
Fig
u
r
e
4
.
R
esNet1
0
1
ar
c
h
itectu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Usi
n
g
R
esN
et
a
r
ch
itectu
r
e
w
it
h
MRI
fo
r
cla
s
s
ifica
tio
n
o
f b
r
a
in
ima
g
es
(
S
u
b
r
a
ma
n
ia
n
Dh
a
n
a
la
ksh
mi
)
153
3
.
5
.
ResNet
1
5
2
T
h
e
b
i
g
g
est
R
esNet
v
ar
iatio
n
with
1
5
2
lay
er
s
.
I
t
m
ay
r
ep
r
esen
t
m
o
r
e
co
m
p
lex
ch
ar
ac
te
r
is
tics
b
u
t
n
ee
d
s
m
o
r
e
p
r
o
ce
s
s
in
g
r
eso
u
r
ce
s
f
o
r
tr
ain
in
g
an
d
in
f
er
en
c
e.
R
esNet
-
1
5
2
,
with
its
1
5
2
-
l
ay
er
d
esig
n
,
is
th
e
ap
ex
o
f
R
esNet
d
ee
p
lea
r
n
in
g
b
r
ea
k
th
r
o
u
g
h
.
T
o
o
v
er
co
m
e
th
e
v
an
is
h
in
g
g
r
ad
ien
t
is
s
u
e,
R
esNet
-
1
5
2
u
s
es
s
ev
er
al
r
esid
u
al
b
lo
ck
s
with
id
en
tity
s
h
o
r
tcu
ts
to
tr
ai
n
d
ee
p
n
etwo
r
k
s
.
E
ac
h
b
lo
ck
h
as
c
o
n
v
o
lu
tio
n
al
lay
er
s
,
b
atch
n
o
r
m
alis
atio
n
,
an
d
R
eL
U
ac
tiv
atio
n
s
f
o
r
tr
ain
in
g
s
tab
ilit
y
an
d
ef
f
icien
c
y
.
B
o
ttlen
ec
k
lay
er
s
o
p
tim
is
e
co
m
p
u
tin
g
r
eso
u
r
ce
s
a
n
d
r
e
d
u
ce
p
ar
am
eter
s
with
o
u
t a
f
f
ec
tin
g
p
er
f
o
r
m
an
ce
.
R
esNet
-
1
5
2
's
d
ep
th
lets
it
ca
t
ch
co
m
p
licated
p
atter
n
s
an
d
f
ea
tu
r
es,
m
ak
in
g
it
i
d
ea
l
f
o
r
s
o
p
h
is
ticated
im
ag
e
class
if
icatio
n
,
o
b
ject
i
d
en
tific
ati
o
n
,
an
d
s
em
an
tic
s
eg
m
en
tatio
n
.
T
h
e
m
o
d
el'
s
ex
t
r
em
e
ac
cu
r
ac
y
an
d
s
ca
lab
ilit
y
h
av
e
m
ad
e
it
a
co
m
p
u
ter
v
is
io
n
b
en
ch
m
a
r
k
,
a
d
v
a
n
cin
g
d
ee
p
lear
n
in
g
d
is
co
v
e
r
ies
an
d
ap
p
licatio
n
s
.
T
h
e
Fig
u
r
e
5
s
h
o
ws
th
e
R
esNet
1
5
2
Ar
c
h
itectu
r
e.
I
n
th
is
th
e
co
n
v
o
lu
tio
n
o
f
7
x
7
o
f
6
4
is
r
ep
ea
ted
o
n
ce
,
an
d
a
f
t
e
r
m
a
x
p
o
o
l
i
n
g
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
1
x
1
o
f
6
4
is
r
e
p
e
a
te
d
3
t
i
m
es
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
3
x
3
o
f
6
4
i
s
r
e
p
ea
t
e
d
3
t
i
m
e
s
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
1
x
1
o
f
128
i
s
r
e
p
e
a
t
e
d
8
t
i
m
e
s
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
3
x
3
o
f
1
2
8
i
s
r
e
p
e
a
t
e
d
8
t
i
m
e
s
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
1
x
1
o
f
256
i
s
r
e
p
e
a
t
e
d
3
9
t
i
m
es
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
3
x
3
o
f
2
5
6
is
r
e
p
e
a
t
e
d
36
ti
m
e
s
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
1
x
1
o
f
512
is
r
e
p
e
a
t
e
d
1
1
t
i
m
es
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
3
x
3
o
f
5
1
2
i
s
r
e
p
e
a
t
e
d
3
t
i
m
es
,
t
h
e
c
o
n
v
o
l
u
t
i
o
n
o
f
1
x
1
o
f
1
0
2
4
i
s
r
e
p
e
a
t
e
d
3
6
t
i
m
es
,
t
h
e
c
o
n
v
o
l
u
t
io
n
o
f
1
x
1
o
f
2
0
4
8
i
s
r
e
p
e
a
t
e
d
3
t
i
m
e
s
.
Fig
u
r
e
5
.
R
esNet1
5
2
ar
c
h
itectu
r
e
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
r
esu
lts
ar
e
s
ee
n
in
th
e
f
r
ee
ly
ac
ce
s
s
ib
le
R
E
M
B
R
A
ND
T
d
atab
ase.
C
u
r
r
en
tly
,
1
3
3
MRI
b
r
ain
im
ag
es,
m
ea
s
u
r
in
g
2
5
6
×
2
5
6
p
ix
els
ea
ch
,
a
r
e
ac
ce
s
s
ib
le.
Af
ter
5
0
0
im
ag
e
g
r
ap
h
s
ar
e
p
ic
k
e
d
f
r
o
m
th
e
n
o
r
m
al
ca
teg
o
r
y
a
n
d
5
0
0
f
r
o
m
th
e
a
b
n
o
r
m
al
ca
teg
o
r
y
,
th
e
t
o
tal
n
u
m
b
er
o
f
im
a
g
es
o
b
tain
ed
f
r
o
m
th
e
d
atab
ase
is
1
0
0
0
.
Fig
u
r
e
6
d
is
p
lay
s
th
e
a
b
er
r
an
t
im
a
g
es
f
r
o
m
th
e
R
E
MBR
A
NDT
d
atab
ase.
T
h
e
k
-
f
o
ld
(
1
0
-
f
o
ld
)
c
r
o
s
s
-
v
alid
atio
n
ap
p
r
o
ac
h
is
u
s
ed
b
y
th
e
s
u
g
g
ested
R
esNet
m
o
d
el.
Fig
u
r
e
7
s
h
o
ws
ex
am
p
les
o
f
n
o
r
m
al
im
ag
es
tak
en
f
r
o
m
th
e
R
E
MB
R
ANDT
d
atab
ase.
T
h
e
k
-
f
o
l
d
(
1
0
-
f
o
ld
)
cr
o
s
s
-
v
alid
atio
n
ap
p
r
o
ac
h
is
u
s
ed
b
y
th
e
s
u
g
g
es
t
ed
R
esNet
m
o
d
el.
W
h
en
it
co
m
es
to
ass
e
s
s
in
g
p
er
f
o
r
m
an
ce
,
th
e
R
esNet
s
y
s
tem
is
em
p
lo
y
in
g
f
o
u
r
p
ar
am
e
ter
s
:
T
P,
T
N,
FP
,
an
d
FN,
w
h
ich
s
tan
d
f
o
r
T
r
u
e
Po
s
itiv
e,
T
r
u
e
Neg
ativ
e
,
s
en
s
itiv
ity
,
an
d
s
p
ec
if
icity
,
r
esp
ec
tiv
ely
.
T
ab
le
1
s
h
o
w
th
e
f
o
r
m
u
la
b
ased
R
esNet
tec
h
n
iq
u
e
o
f
T
P,
T
N,
FP
an
d
FN
f
o
r
s
en
s
itiv
ity
,
ac
cu
r
ac
y
,
s
p
ec
if
icity
.
T
h
e
R
esNet
d
ata
b
ases
ar
e
s
h
o
wn
in
T
ab
les 2
to
5
,
wh
er
e
T
P,
T
N,
FP
,
an
d
FN a
r
e
th
e
v
alu
es u
s
e
d
.
Fig
u
r
e
6
.
I
l
u
s
tr
atio
n
s
o
f
ab
n
o
r
m
al
b
r
ain
im
ag
es f
r
o
m
t
h
e
R
E
MBR
A
NDT
d
atab
ase
u
s
ed
b
y
R
esNet
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
1
48
-
1
58
154
R
esNet
1
8
,
R
e
s
Net
3
4
,
an
d
R
e
s
Net
5
0
,
wh
ich
ar
e
v
er
s
io
n
s
o
f
th
e
R
esid
u
al
Neu
r
al
Netwo
r
k
(
R
esNet)
ar
ch
itectu
r
e.
R
esNet
1
8
's
r
ap
id
tr
ain
in
g
with
m
o
d
est
m
e
m
o
r
y
u
tili
za
tio
n
m
ak
es
it
a
g
o
o
d
f
it
f
o
r
s
m
aller
d
atasets
.
W
ith
its
en
h
an
ce
d
d
ep
th
,
R
esNet
3
4
im
p
r
o
v
es
u
p
o
n
R
esNet
1
8
,
m
ak
in
g
it
m
o
r
e
ef
f
icien
t
an
d
b
ette
r
ab
le
to
g
en
er
alize
o
v
er
b
i
g
g
er
d
atasets
.
W
h
en
it
co
m
es
to
lar
g
e
-
s
ca
le
d
atasets
,
th
e
d
ee
p
est
m
o
d
el,
R
esNet
5
0
,
o
f
f
er
s
th
e
b
est
f
ea
tu
r
e
ex
tr
ac
t
i
o
n
an
d
r
ep
r
esen
tatio
n
lear
n
in
g
.
On
th
e
o
th
e
r
h
an
d
,
g
r
ea
ter
c
o
m
p
u
tin
g
r
eso
u
r
ce
s
ar
e
n
ee
d
ed
f
o
r
it.
Dep
en
d
i
n
g
o
n
th
e
co
m
p
lex
ity
an
d
am
o
u
n
t
o
f
th
e
d
ataset,
d
if
f
er
en
t
R
esNet
v
ar
ian
ts
p
r
o
v
i
d
e
d
if
f
er
en
t
b
e
n
ef
its
in
im
ag
e
c
lass
if
icatio
n
,
o
b
ject
r
ec
o
g
n
iti
o
n
,
an
d
s
em
an
tic
s
eg
m
e
n
tatio
n
.
T
ab
le
6
s
h
o
ws
R
esNet
ar
ch
itectu
r
al
co
m
p
o
n
e
n
ts
an
d
th
eir
r
esp
o
n
s
ib
ilit
ies,
ad
v
an
tag
es,
f
u
n
ctio
n
s
,
an
d
s
co
p
e
in
b
r
ain
im
ag
e
class
if
icatio
n
f
r
o
m
MRI
s
ca
n
s
.
B
y
p
r
o
v
id
in
g
d
ir
ec
t
g
r
ad
ie
n
t
ch
a
n
n
els
to
o
v
e
r
co
m
e
th
e
v
an
is
h
in
g
g
r
ad
ie
n
t
is
s
u
e,
r
esid
u
al
b
lo
c
k
s
en
ab
l
e
d
ee
p
e
r
n
etwo
r
k
tr
ain
in
g
.
Featu
r
e
ex
tr
ac
tio
n
f
r
o
m
MRI
im
ag
es
r
eq
u
ir
es
co
n
v
o
l
u
tio
n
al
lay
er
s
to
ca
p
tu
r
e
s
p
atial
h
ier
ar
ch
ies.
No
r
m
ali
s
in
g
lay
er
in
p
u
ts
v
ia
b
atc
h
n
o
r
m
alis
atio
n
r
ed
u
ce
s
in
ter
n
al
co
v
ar
iate
s
h
if
t,
im
p
r
o
v
in
g
tr
ai
n
in
g
s
tab
ilit
y
an
d
ef
f
icien
cy
.
R
eL
U
ac
tiv
atio
n
s
p
r
o
v
id
e
n
o
n
-
lin
ea
r
ity
,
en
ab
lin
g
th
e
m
o
d
el
to
lear
n
c
o
m
p
licated
b
r
ain
p
atter
n
s
.
B
o
ttlen
ec
k
lay
er
s
b
alan
c
e
d
ep
t
h
a
n
d
p
e
r
f
o
r
m
an
ce
b
y
d
ec
r
ea
s
in
g
p
ar
am
eter
s
to
o
p
ti
m
is
e
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
I
d
en
tity
s
h
o
r
tcu
ts
av
o
id
g
r
ad
ien
t
f
lo
w
s
lo
wd
o
wn
as
n
etwo
r
k
d
ep
th
g
r
o
ws.
Fin
ally
,
th
e
class
if
icatio
n
lay
er
g
iv
es
lik
elih
o
o
d
v
alu
es
to
class
es
,
ess
en
tial
f
o
r
b
r
ain
d
is
ea
s
e
d
iag
n
o
s
is
.
T
h
ese
co
m
p
o
n
en
ts
ef
f
icien
tly
class
if
y
b
r
ai
n
p
ictu
r
es f
o
r
im
p
r
o
v
ed
m
ed
ic
al
d
iag
n
o
s
is
.
Fig
u
r
e
7
.
I
ll
u
s
tr
atio
n
s
o
f
n
o
r
m
al
b
r
ain
im
ag
es
f
r
o
m
th
e
R
E
MBR
A
NDT
d
atab
ase
u
s
ed
b
y
R
esNet
T
ab
le
1
.
Sen
s
itiv
ity
,
ac
cu
r
ac
y
,
an
d
s
p
ec
if
icity
ar
e
t
h
e
p
er
f
o
r
m
an
ce
m
etr
ics th
at
ar
e
b
ein
g
m
ea
s
u
r
ed
b
y
th
e
r
esn
et
s
y
s
tem
M
e
t
r
i
c
I
mp
o
r
t
a
n
c
e
F
o
r
mu
l
a
S
e
n
s
i
t
i
v
i
t
y
I
n
d
i
c
a
t
e
s
h
o
w
w
e
l
l
t
h
e
t
e
st
c
a
p
t
u
r
e
s
t
h
e
p
r
e
se
n
c
e
o
f
t
h
e
c
o
n
d
i
t
i
o
n
o
r
e
v
e
n
t
o
f
i
n
t
e
r
e
s
t
.
+
A
c
c
u
r
a
c
y
P
r
o
v
i
d
e
s a
g
e
n
e
r
a
l
a
ss
e
ssm
e
n
t
o
f
t
h
e
r
e
l
i
a
b
i
l
i
t
y
a
n
d
p
r
e
c
i
si
o
n
o
f
t
h
e
d
i
a
g
n
o
s
t
i
c
t
o
o
l
o
r
p
r
e
d
i
c
t
i
v
e
a
l
g
o
r
i
t
h
m.
+
+
+
+
S
p
e
c
i
f
i
c
i
t
y
S
h
o
w
s
h
o
w
w
e
l
l
t
h
e
t
e
s
t
a
v
o
i
d
s f
a
l
s
e
a
l
a
r
ms
o
r
i
n
c
o
r
r
e
c
t
l
y
c
l
a
ss
i
f
y
i
n
g
n
e
g
a
t
i
v
e
i
n
s
t
a
n
c
e
s a
s
p
o
s
i
t
i
v
e
.
+
T
ab
le
2
.
Per
f
o
r
m
an
ce
o
f
R
esNet
tech
n
iq
u
e
-
I
A
n
t
i
c
i
p
a
t
e
d
C
l
a
s
s
A
b
n
o
r
m
a
l
n
o
r
m
a
l
R
e
a
l
c
l
a
ss
A
b
n
o
r
m
a
l
4
7
3
(
TP)
2
7
(
F
N
)
n
o
r
m
a
l
2
7
(
F
P
)
4
7
4
(
TN
)
T
ab
le
3
.
Per
f
o
r
m
an
ce
o
f
R
esNet
tech
n
iq
u
e
–
II
A
n
t
i
c
i
p
a
t
e
d
C
l
a
s
s
A
b
n
o
r
m
a
l
n
o
r
m
a
l
R
e
a
l
c
l
a
ss
A
b
n
o
r
m
a
l
4
7
8
(
TP)
2
1
(
F
N
)
n
o
r
m
a
l
2
2
(
F
P
)
4
7
9
(
TN
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Usi
n
g
R
esN
et
a
r
ch
itectu
r
e
w
it
h
MRI
fo
r
cla
s
s
ifica
tio
n
o
f b
r
a
in
ima
g
es
(
S
u
b
r
a
ma
n
ia
n
Dh
a
n
a
la
ksh
mi
)
155
T
ab
le
4
.
Per
f
o
r
m
an
ce
o
f
R
esNet
tech
n
iq
u
e
-
III
A
n
t
i
c
i
p
a
t
e
d
C
l
a
s
s
A
b
n
o
r
m
a
l
n
o
r
m
a
l
R
e
a
l
c
l
a
ss
A
b
n
o
r
m
a
l
4
8
0
(
TP)
1
8
(
F
N
)
n
o
r
m
a
l
2
0
(FP)
4
8
2
(
TN
)
T
ab
le
5
.
Per
f
o
r
m
an
ce
o
f
R
esNet
tech
n
iq
u
e
-
IV
A
n
t
i
c
i
p
a
t
e
d
C
l
a
s
s
A
b
n
o
r
m
a
l
n
o
r
m
a
l
R
e
a
l
c
l
a
ss
A
b
n
o
r
m
a
l
4
7
1
(
TP)
3
2
(
F
N
)
n
o
r
m
a
l
2
9
(FP)
4
6
8
(
TN
)
T
ab
le
7
lis
ts
th
e
s
tr
en
g
th
s
an
d
lim
itatio
n
s
o
f
R
esNe
t
ar
ch
itectu
r
e
f
o
r
MRI
b
r
ain
im
a
g
e
class
if
icatio
n
.
R
esid
u
al
b
lo
ck
s
allo
w
d
ee
p
n
etwo
r
k
tr
ain
in
g
b
u
t
ad
d
c
o
m
p
u
tatio
n
al
co
m
p
le
x
ity
.
C
o
n
v
o
l
u
tio
n
al
lay
er
s
n
ee
d
p
len
ty
o
f
m
em
o
r
y
an
d
co
m
p
u
te
to
r
ec
o
r
d
s
p
atial
h
ier
ar
ch
ies.
B
atch
n
o
r
m
alis
atio
n
s
p
ee
d
s
u
p
tr
ain
in
g
b
u
t
in
cr
ea
s
es
co
m
p
u
tatio
n
al
lo
ad
.
C
o
m
p
lex
p
atter
n
s
ca
n
b
e
le
ar
n
ed
b
y
R
eL
U
ac
tiv
atio
n
s
,
b
u
t
th
ey
m
ay
k
ill
n
eu
r
o
n
s
.
I
n
b
o
ttlen
ec
k
lay
e
r
s
,
p
ar
am
eter
s
a
r
e
c
o
m
p
r
ess
ed
,
s
o
m
etim
es
lo
s
in
g
in
f
o
r
m
atio
n
.
I
d
e
n
tity
s
h
o
r
tc
u
ts
s
im
p
lify
g
r
ad
ien
t
f
lo
w
b
u
t
co
m
p
licate
n
etwo
r
k
d
esig
n
.
Fin
a
lly
,
th
e
class
if
icatio
n
lay
er
ap
p
r
o
p
r
iately
p
r
o
v
i
d
es
p
r
o
b
a
b
ilit
y
v
alu
es
f
o
r
class
es,
alth
o
u
g
h
it
m
a
y
o
v
er
f
it
if
n
o
t
r
eg
u
lar
is
ed
.
T
h
ese
f
ac
to
r
s
e
n
ab
le
ef
f
icien
t
y
et
r
eso
u
r
ce
-
in
ten
s
iv
e
b
r
ain
MRI
p
ictu
r
e
ca
teg
o
r
izatio
n
.
Fig
u
r
e
8
d
is
p
lay
s
th
e
ca
lcu
l
ated
p
er
f
o
r
m
a
n
ce
m
etr
ics
b
ased
o
n
th
e
R
esNet
T
ec
h
n
iq
u
e.
W
ith
an
o
v
er
all
ac
cu
r
ac
y
o
f
9
6
.
2
%,
s
en
s
itiv
ity
o
f
9
6
%,
an
d
s
p
ec
if
icity
o
f
9
6
.
4
%,
R
esNet
f
ea
tu
r
es c
lea
r
ly
o
u
tp
e
r
f
o
r
m
th
e
co
m
p
etitio
n
i
s
s
h
o
wn
in
Fig
u
r
e
8
.
T
ab
le
6
.
Asp
ec
ts
o
f
r
esn
et
a
r
c
h
itectu
r
e
W
ith
MRI
f
o
r
b
r
ai
n
i
m
ag
e
class
if
icatio
n
A
sp
e
c
t
R
o
l
e
B
e
n
e
f
i
t
F
u
n
c
t
i
o
n
S
c
o
p
e
R
e
si
d
u
a
l
B
l
o
c
k
s
En
a
b
l
e
d
e
e
p
e
r
n
e
t
w
o
r
k
t
r
a
i
n
i
n
g
M
i
t
i
g
a
t
e
s
v
a
n
i
sh
i
n
g
g
r
a
d
i
e
n
t
p
r
o
b
l
e
m
A
l
l
o
w
s
d
i
r
e
c
t
p
a
t
h
s fo
r
g
r
a
d
i
e
n
t
f
l
o
w
Essen
t
i
a
l
f
o
r
b
u
i
l
d
i
n
g
v
e
r
y
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
C
o
n
v
o
l
u
t
i
o
n
a
l
La
y
e
r
s
Ex
t
r
a
c
t
f
e
a
t
u
r
e
s fr
o
m
M
R
I
i
ma
g
e
s
C
a
p
t
u
r
e
s
d
e
t
a
i
l
e
d
sp
a
t
i
a
l
h
i
e
r
a
r
c
h
i
e
s
P
r
o
c
e
ss
e
s raw
i
m
a
g
e
d
a
t
a
i
n
t
o
me
a
n
i
n
g
f
u
l
f
e
a
t
u
r
e
s
C
r
u
c
i
a
l
f
o
r
r
e
c
o
g
n
i
z
i
n
g
c
o
m
p
l
e
x
p
a
t
t
e
r
n
s
i
n
b
r
a
i
n
i
ma
g
e
s
B
a
t
c
h
N
o
r
mal
i
z
a
t
i
o
n
S
t
a
b
i
l
i
z
e
s
a
n
d
a
c
c
e
l
e
r
a
t
e
s t
r
a
i
n
i
n
g
R
e
d
u
c
e
s
i
n
t
e
r
n
a
l
c
o
v
a
r
i
a
t
e
s
h
i
f
t
N
o
r
mal
i
z
e
s l
a
y
e
r
i
n
p
u
t
s
En
h
a
n
c
e
s t
r
a
i
n
i
n
g
e
f
f
i
c
i
e
n
c
y
a
n
d
c
o
n
v
e
r
g
e
n
c
e
sp
e
e
d
R
e
LU
A
c
t
i
v
a
t
i
o
n
s
I
n
t
r
o
d
u
c
e
s
n
o
n
-
l
i
n
e
a
r
i
t
y
A
l
l
o
w
s
m
o
d
e
l
t
o
l
e
a
r
n
c
o
m
p
l
e
x
p
a
t
t
e
r
n
s
A
p
p
l
i
e
s
n
o
n
-
l
i
n
e
a
r
t
r
a
n
sf
o
r
ma
t
i
o
n
I
mp
o
r
t
a
n
t
f
o
r
c
a
p
t
u
r
i
n
g
i
n
t
r
i
c
a
t
e
d
e
t
a
i
l
s
i
n
M
R
I
d
a
t
a
B
o
t
t
l
e
n
e
c
k
La
y
e
r
s
O
p
t
i
mi
z
e
c
o
m
p
u
t
a
t
i
o
n
a
l
e
f
f
i
c
i
e
n
c
y
R
e
d
u
c
e
s
n
u
m
b
e
r
o
f
p
a
r
a
m
e
t
e
r
s
C
o
m
b
i
n
e
s m
u
l
t
i
p
l
e
l
a
y
e
r
s
i
n
t
o
a
c
o
mp
a
c
t
f
o
r
m
B
a
l
a
n
c
e
s
d
e
p
t
h
a
n
d
p
e
r
f
o
r
m
a
n
c
e
i
n
r
e
s
o
u
r
c
e
-
c
o
n
st
r
a
i
n
e
d
se
t
t
i
n
g
s
I
d
e
n
t
i
t
y
S
h
o
r
t
c
u
t
s
F
a
c
i
l
i
t
a
t
e
g
r
a
d
i
e
n
t
f
l
o
w
P
r
e
v
e
n
t
s
d
e
g
r
a
d
a
t
i
o
n
o
f
d
e
e
p
n
e
t
w
o
r
k
s
A
d
d
s
sh
o
r
t
c
u
t
c
o
n
n
e
c
t
i
o
n
s wi
t
h
i
n
r
e
si
d
u
a
l
b
l
o
c
k
s
K
e
y
f
o
r
ma
i
n
t
a
i
n
i
n
g
p
e
r
f
o
r
m
a
n
c
e
a
s
n
e
t
w
o
r
k
d
e
p
t
h
i
n
c
r
e
a
s
e
s
C
l
a
s
si
f
i
c
a
t
i
o
n
La
y
e
r
D
e
t
e
r
m
i
n
e
s
o
u
t
p
u
t
c
a
t
e
g
o
r
y
A
ssi
g
n
s
p
r
o
b
a
b
i
l
i
t
y
sco
r
e
s
t
o
d
i
f
f
e
r
e
n
t
c
l
a
ss
e
s
P
r
o
c
e
ss
e
s
e
x
t
r
a
c
t
e
d
f
e
a
t
u
r
e
s
i
n
t
o
c
l
a
ss
i
f
i
c
a
t
i
o
n
V
i
t
a
l
f
o
r
d
i
a
g
n
o
s
i
n
g
b
r
a
i
n
c
o
n
d
i
t
i
o
n
s
b
a
se
d
o
n
M
R
I
sca
n
s
T
ab
le
7
.
Pro
s
an
d
co
n
s
o
f
R
esNet
ar
ch
itectu
r
e
wi
th
MRI
f
o
r
b
r
ain
im
ag
e
c
lass
if
icatio
n
A
sp
e
c
t
P
r
o
s
C
o
n
s
R
e
si
d
u
a
l
B
l
o
c
k
s
En
a
b
l
e
t
r
a
i
n
i
n
g
o
f
v
e
r
y
d
e
e
p
n
e
t
w
o
r
k
s
I
n
c
r
e
a
se
d
c
o
m
p
u
t
a
t
i
o
n
a
l
c
o
m
p
l
e
x
i
t
y
C
o
n
v
o
l
u
t
i
o
n
a
l
L
a
y
e
r
s
C
a
p
t
u
r
e
d
e
t
a
i
l
e
d
sp
a
t
i
a
l
h
i
e
r
a
r
c
h
i
e
s
H
i
g
h
c
o
m
p
u
t
a
t
i
o
n
a
l
a
n
d
m
e
mo
r
y
r
e
q
u
i
r
e
me
n
t
s
B
a
t
c
h
N
o
r
m
a
l
i
z
a
t
i
o
n
S
t
a
b
i
l
i
z
e
a
n
d
a
c
c
e
l
e
r
a
t
e
t
r
a
i
n
i
n
g
A
d
d
i
t
i
o
n
a
l
c
o
m
p
u
t
a
t
i
o
n
a
l
o
v
e
r
h
e
a
d
R
e
LU
A
c
t
i
v
a
t
i
o
n
s
A
l
l
o
w
l
e
a
r
n
i
n
g
o
f
c
o
mp
l
e
x
p
a
t
t
e
r
n
s
P
o
t
e
n
t
i
a
l
f
o
r
d
e
a
d
n
e
u
r
o
n
s
(
n
e
u
r
o
n
s
n
o
t
a
c
t
i
v
a
t
i
n
g
)
B
o
t
t
l
e
n
e
c
k
La
y
e
r
s
R
e
d
u
c
e
n
u
m
b
e
r
o
f
p
a
r
a
m
e
t
e
r
s
P
o
ssi
b
l
e
l
o
ss
o
f
i
n
f
o
r
ma
t
i
o
n
t
h
r
o
u
g
h
c
o
mp
r
e
ssi
o
n
I
d
e
n
t
i
t
y
S
h
o
r
t
c
u
t
s
F
a
c
i
l
i
t
a
t
e
sm
o
o
t
h
g
r
a
d
i
e
n
t
f
l
o
w
M
a
y
a
d
d
c
o
m
p
l
e
x
i
t
y
t
o
n
e
t
w
o
r
k
d
e
si
g
n
C
l
a
s
si
f
i
c
a
t
i
o
n
La
y
e
r
A
ssi
g
n
a
c
c
u
r
a
t
e
p
r
o
b
a
b
i
l
i
t
y
s
c
o
r
e
s fo
r
c
l
a
ss
e
s
S
u
sce
p
t
i
b
l
e
t
o
o
v
e
r
f
i
t
t
i
n
g
i
f
n
o
t
p
r
o
p
e
r
l
y
r
e
g
u
l
a
r
i
z
e
d
Fig
u
r
e
8
d
is
p
lay
s
th
e
ca
lcu
la
ted
p
er
f
o
r
m
an
ce
m
etr
ics
b
as
ed
o
n
th
e
R
esNet
T
ec
h
n
iq
u
e
.
W
ith
an
o
v
er
all
ac
cu
r
ac
y
o
f
9
6
.
2
%,
s
en
s
itiv
ity
o
f
9
6
%,
an
d
s
p
ec
if
icity
o
f
9
6
.
4
%
.
Fin
ally
,
th
e
class
if
icatio
n
lay
er
ap
p
r
o
p
r
iately
p
r
o
v
id
es
p
r
o
b
ab
ilit
y
v
alu
es
f
o
r
class
es,
a
lth
o
u
g
h
it
m
ay
o
v
er
f
it
if
n
o
t
r
e
g
u
la
r
is
ed
.
T
h
ese
f
ac
to
r
s
en
ab
le
ef
f
icien
t
y
et
r
eso
u
r
ce
-
in
ten
s
iv
e
b
r
ain
MRI
p
ictu
r
e
ca
t
eg
o
r
izatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
1
48
-
1
58
156
Fig
u
r
e
8
.
E
v
alu
atio
n
c
r
iter
ia
f
o
r
th
e
s
u
g
g
ested
R
esNet
s
y
s
te
m
in
th
e
d
ia
g
n
o
s
is
o
f
b
r
ain
ca
n
ce
r
5.
CO
NCLU
SI
O
N
Sig
n
if
ican
t
ad
v
an
ce
s
in
m
ed
i
ca
l
d
iag
n
o
s
is
m
ay
b
e
m
ad
e
v
ia
th
e
u
s
e
o
f
R
e
s
Net
m
eth
o
d
s
f
o
r
b
r
ain
im
ag
e
ca
teg
o
r
izatio
n
in
MRI.
Par
ticu
lar
ly
f
o
r
u
n
co
m
m
o
n
n
e
u
r
o
lo
g
ical
d
is
o
r
d
er
s
,
th
e
n
ee
d
f
o
r
b
ig
an
d
v
ar
ied
d
atasets
to
tr
ain
s
tr
o
n
g
m
o
d
el
s
co
n
tin
u
es
to
b
e
an
o
b
s
tacle
.
Qu
esti
o
n
s
o
f
tr
u
s
t
wo
r
th
in
ess
an
d
d
ep
en
d
a
b
ilit
y
ar
e
b
r
o
u
g
h
t u
p
b
y
th
e
in
ter
p
r
et
ab
ilit
y
o
f
d
ee
p
lear
n
in
g
m
o
d
el
s
wh
en
u
s
ed
in
h
ea
lth
ca
r
e
co
n
tex
ts
.
I
t w
ill n
ee
d
a
team
ef
f
o
r
t
f
r
o
m
c
o
m
p
u
ter
s
cien
tis
ts
,
r
ad
io
lo
g
is
ts
,
an
d
o
th
er
m
ed
ical
ex
p
e
r
ts
to
o
v
er
c
o
m
e
th
ese
o
b
s
tacle
s
.
T
h
e
p
o
te
n
tial
f
u
tu
r
e
r
a
m
if
ica
tio
n
s
o
f
t
h
is
s
tu
d
y
ar
e
en
c
o
u
r
ag
in
g
.
T
o
f
u
r
th
er
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
a
n
d
in
ter
p
r
etab
ilit
y
o
f
MRI
d
ata,
f
u
tu
r
e
s
tu
d
ies
m
ay
in
v
esti
g
ate
n
ew
d
ee
p
lear
n
in
g
a
r
ch
itect
u
r
es
d
ev
elo
p
ed
f
o
r
th
is
k
in
d
o
f
d
ata.
A
m
o
r
e
c
o
m
p
lete
im
ag
e
o
f
b
r
ain
h
ea
l
t
h
m
ig
h
t
b
e
s
h
o
wn
b
y
im
p
r
o
v
in
g
class
if
icatio
n
p
er
f
o
r
m
an
ce
a
n
d
in
te
g
r
atin
g
m
u
ltimo
d
al
d
ata
s
o
u
r
ce
s
lik
e
d
if
f
u
s
io
n
te
n
s
o
r
im
ag
i
n
g
a
n
d
f
u
n
ctio
n
al
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
.
R
ec
en
t
d
ev
elo
p
m
en
ts
in
p
r
i
v
ac
y
-
p
r
eser
v
in
g
ap
p
r
o
ac
h
es
an
d
f
ed
e
r
a
ted
lear
n
in
g
h
a
v
e
o
p
en
e
d
n
ew
p
o
s
s
ib
ilit
ies
f
o
r
t
r
ain
in
g
m
o
d
els
u
s
in
g
d
ec
en
tr
a
lized
d
ata
w
h
ile
s
till
p
r
o
tectin
g
p
atien
ts
'
p
r
iv
ac
y
.
T
h
er
e
is
tr
em
en
d
o
u
s
h
o
p
e
th
a
t
f
u
r
th
er
d
ev
el
o
p
m
en
t
o
f
R
esNet
m
eth
o
d
s
f
o
r
b
r
ain
im
a
g
e
ca
teg
o
r
izatio
n
m
ay
ch
an
g
e
clin
ical
p
r
ac
tice
b
y
all
o
win
g
f
o
r
th
e
e
a
r
lier
d
iag
n
o
s
i
s
an
d
m
o
r
e
tailo
r
ed
tr
ea
tm
en
t
o
f
a
wid
e
r
an
g
e
o
f
n
eu
r
o
lo
g
ical
d
is
ea
s
es.
Re
su
lt
s
p
ro
v
e
d
t
h
a
t
t
h
e
p
r
o
p
o
se
d
Re
sN
e
t
sy
ste
m
a
c
h
iev
e
s
9
8
.
8
%
o
v
e
ra
ll
a
c
c
u
ra
c
y
with
9
8
.
6
%
se
n
siti
v
it
y
a
n
d
9
9
%
sp
e
c
ifi
c
it
y
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Fu
n
d
in
g
in
f
o
r
m
ati
o
n
is
n
o
t a
v
ailab
le.
AUTHO
R
CO
NTR
IBU
TIO
N
S
S
TATEMENT
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Su
b
r
am
an
ian
Dh
an
alak
s
h
m
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Su
b
r
am
an
ian
Ar
u
ls
elv
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
T
h
e
au
th
o
r
s
h
av
e
n
o
co
n
f
lict o
f
in
ter
est r
elev
an
t to
t
h
is
p
ap
er
.
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
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
aila
b
le
o
n
r
eq
u
est
f
r
o
m
th
e
co
r
r
esp
o
n
d
in
g
au
th
o
r
,
[
S.
D]
.
T
h
e
d
ata,
wh
ich
co
n
tain
in
f
o
r
m
atio
n
th
at
co
u
ld
co
m
p
r
o
m
is
e
th
e
p
r
iv
ac
y
o
f
r
esear
c
h
p
ar
ticip
an
ts
,
ar
e
n
o
t p
u
b
licly
a
v
ailab
le
d
u
e
to
ce
r
tain
r
estrictio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Usi
n
g
R
esN
et
a
r
ch
itectu
r
e
w
it
h
MRI
fo
r
cla
s
s
ifica
tio
n
o
f b
r
a
in
ima
g
es
(
S
u
b
r
a
ma
n
ia
n
Dh
a
n
a
la
ksh
mi
)
157
RE
F
E
R
E
NC
E
S
[
1
]
S
.
K
a
r
t
h
i
k
a
,
S
.
D
h
a
n
a
l
a
k
s
h
mi
,
S
.
M
.
M
u
r
t
h
y
,
N
.
M
i
s
h
r
a
,
S
.
S
a
si
k
a
l
a
,
a
n
d
S
.
M
u
r
u
g
a
n
,
“
R
a
s
p
b
e
r
r
y
P
i
-
e
n
a
b
l
e
d
w
e
a
r
a
b
l
e
s
e
n
so
r
s f
o
r
p
e
r
s
o
n
a
l
h
e
a
l
t
h
t
r
a
c
k
i
n
g
a
n
d
a
n
a
l
y
si
s
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
e
l
f
-
S
u
st
a
i
n
i
n
g
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
S
y
s
t
e
m
s (I
C
S
S
A
S
)
,
p
p
.
1
2
4
9
–
1
2
5
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
S
A
S
5
7
9
1
8
.
2
0
2
3
.
1
0
3
3
1
9
0
9
.
[
2
]
A
.
D
e
e
p
a
,
N
.
K
.
M
a
n
i
k
a
n
d
a
n
,
R
.
La
t
h
a
,
J.
P
r
e
e
t
h
a
,
T
.
S
.
K
u
mar,
a
n
d
S
.
M
u
r
u
g
a
n
,
“
I
o
T
-
b
a
se
d
w
e
a
r
a
b
l
e
d
e
v
i
c
e
s fo
r
p
e
r
s
o
n
a
l
s
a
f
e
t
y
a
n
d
a
c
c
i
d
e
n
t
p
r
e
v
e
n
t
i
o
n
sy
s
t
e
ms
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
S
e
c
o
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
m
a
r
t
T
e
c
h
n
o
l
o
g
i
e
s
f
o
r
S
m
a
rt
N
a
t
i
o
n
(
S
m
a
rt
T
e
c
h
C
o
n
)
,
p
p
.
1
5
1
0
–
1
5
1
4
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
S
mart
Te
c
h
C
o
n
5
7
5
2
6
.
2
0
2
3
.
1
0
3
9
1
6
9
1
.
[
3
]
R
.
K
.
V
a
n
a
k
a
mam
i
d
i
,
L
.
R
a
ma
l
i
n
g
a
m,
N
.
A
b
i
r
a
mi
,
S
.
P
r
i
y
a
n
k
a
,
C
.
S
.
K
u
mar,
a
n
d
S
.
M
u
r
u
g
a
n
,
“
I
o
T
sec
u
r
i
t
y
b
a
s
e
d
o
n
mac
h
i
n
e
l
e
a
r
n
i
n
g
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
S
e
c
o
n
d
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
S
m
a
r
t
T
e
c
h
n
o
l
o
g
i
e
s
f
o
r
S
m
a
rt
N
a
t
i
o
n
(
S
m
a
r
t
T
e
c
h
C
o
n
)
,
p
p
.
6
8
3
–
6
8
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
S
ma
r
t
Te
c
h
C
o
n
5
7
5
2
6
.
2
0
2
3
.
1
0
3
9
1
7
2
7
.
[
4
]
M
.
S
.
K
u
mar
,
H
.
A
z
a
t
h
,
A
.
K
.
V
e
l
m
u
r
u
g
a
n
,
K
.
P
a
d
ma
n
a
b
a
n
,
a
n
d
M
.
S
u
b
b
i
a
h
,
“
P
r
e
d
i
c
t
i
o
n
o
f
A
l
z
h
e
i
mer'
s
d
i
s
e
a
se
u
si
n
g
h
y
b
r
i
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
,
”
AI
P
C
o
n
f
e
re
n
c
e
Pr
o
c
e
e
d
i
n
g
s
,
v
o
l
.
2
5
2
3
,
n
o
.
1
,
p
p
.
1
–
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
6
3
/
5
.
0
1
1
0
2
8
3
.
[
5
]
R
.
R
a
ma
n
,
K
.
D
h
i
v
y
a
,
P
.
S
a
p
r
a
,
S
.
G
u
r
p
u
r
,
S
.
P
.
M
a
n
i
r
a
j
,
a
n
d
S
.
M
u
r
u
g
a
n
,
“
I
o
T
-
d
r
i
v
e
n
sm
a
r
t
p
a
c
k
a
g
i
n
g
f
o
r
p
h
a
r
mac
e
u
t
i
c
a
l
s:
En
s
u
r
i
n
g
p
r
o
d
u
c
t
i
n
t
e
g
r
i
t
y
a
n
d
p
a
t
i
e
n
t
saf
e
t
y
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
f
o
r
I
n
n
o
v
a
t
i
o
n
s
i
n
H
e
a
l
t
h
c
a
re
I
n
d
u
s
t
ri
e
s (
I
C
AI
I
H
I
)
,
p
p
.
1
–
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
I
I
H
I
5
7
8
7
1
.
2
0
2
3
.
1
0
4
8
9
4
2
0
.
[
6
]
K
.
P
a
d
ma
n
a
b
a
n
,
A
.
M
.
S
.
K
u
mar,
H
.
A
z
a
t
h
,
A
.
K
.
V
e
l
m
u
r
u
g
a
n
,
a
n
d
M
.
S
u
b
b
i
a
h
,
“
H
y
b
r
i
d
d
a
t
a
m
i
n
i
n
g
t
e
c
h
n
i
q
u
e
b
a
s
e
d
b
r
e
a
st
c
a
n
c
e
r
p
r
e
d
i
c
t
i
o
n
,
”
AI
P
C
o
n
f
e
re
n
c
e
P
ro
c
e
e
d
i
n
g
s
,
v
o
l
.
2
5
2
3
,
n
o
.
1
,
p
p
.
1
–
7
,
2
0
2
3
.
[
7
]
B
.
J
.
G
a
n
e
s
h
,
P
.
V
i
j
a
y
a
n
,
V
.
V
a
i
d
e
h
i
,
S
.
M
u
r
u
g
a
n
,
R
.
M
e
e
n
a
k
s
h
i
,
a
n
d
M
.
R
a
j
m
o
h
a
n
,
“
S
V
M
-
b
a
s
e
d
p
r
e
d
i
c
t
i
v
e
m
o
d
e
l
i
n
g
o
f
d
r
o
w
s
i
n
e
ss
i
n
h
o
sp
i
t
a
l
s
t
a
f
f
f
o
r
o
c
c
u
p
a
t
i
o
n
a
l
s
a
f
e
t
y
s
o
l
u
t
i
o
n
v
i
a
I
o
T
i
n
f
r
a
st
r
u
c
t
u
r
e
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
n
d
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
i
n
g
,
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
C
o
n
t
r
o
l
(
I
C
)
,
p
p
.
1
–
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
4
5
7
4
3
4
.
2
0
2
4
.
1
0
4
8
6
4
2
9
.
[
8
]
M
.
M
.
M
u
st
h
a
f
a
,
T.
R
.
M
a
h
e
s
h
,
V
.
V
i
n
o
t
h
K
u
mar
,
a
n
d
S
.
G
u
l
u
w
a
d
i
,
“
E
n
h
a
n
c
i
n
g
b
r
a
i
n
t
u
m
o
r
d
e
t
e
c
t
i
o
n
i
n
M
R
I
i
ma
g
e
s
t
h
r
o
u
g
h
e
x
p
l
a
i
n
a
b
l
e
A
I
u
si
n
g
G
r
a
d
-
C
A
M
w
i
t
h
R
e
sN
e
t
-
5
0
,
”
B
MC
Me
d
i
c
a
l
I
m
a
g
i
n
g
,
v
o
l
.
2
4
,
n
o
.
1
,
p
p
.
1
–
1
9
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s
1
2
8
8
0
-
0
2
4
-
0
1
2
9
2
-
7.
[
9
]
A
.
A
g
h
a
e
i
a
n
d
M
.
E.
M
o
g
h
a
d
d
a
m,
“
B
r
a
i
n
a
g
e
g
a
p
e
s
t
i
m
a
t
i
o
n
u
si
n
g
a
t
t
e
n
t
i
o
n
-
b
a
s
e
d
R
e
sN
e
t
me
t
h
o
d
f
o
r
A
l
z
h
e
i
m
e
r
’
s
d
i
s
e
a
se
d
e
t
e
c
t
i
o
n
,
”
Br
a
i
n
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
1
1
,
n
o
.
1
,
p
p
.
1
–
1
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s
4
0
7
0
8
-
0
2
4
-
0
0
2
3
0
-
1.
[
1
0
]
W
.
Y
a
n
g
,
“
B
r
a
i
n
r
e
s
o
n
a
n
c
e
i
m
a
g
i
n
g
seg
m
e
n
t
a
t
i
o
n
me
t
h
o
d
b
a
se
d
o
n
d
e
e
p
R
e
sU
N
e
t
,
”
Ad
v
a
n
c
e
s
i
n
En
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
Re
se
a
rc
h
,
v
o
l
.
1
0
,
n
o
.
1
,
p
p
.
4
4
8
–
4
5
4
,
2
0
2
4
,
d
o
i
:
1
0
.
5
6
0
2
8
/
a
e
t
r
.
1
0
.
1
.
4
4
8
.
2
0
2
4
.
[
1
1
]
N
.
El
a
z
a
b
,
W
.
A
.
G
a
b
-
A
l
l
a
h
,
a
n
d
M
.
El
m
o
g
y
,
“
A
m
u
l
t
i
-
c
l
a
ss
b
r
a
i
n
t
u
m
o
r
g
r
a
d
i
n
g
s
y
st
e
m
b
a
se
d
o
n
h
i
st
o
p
a
t
h
o
l
o
g
i
c
a
l
i
ma
g
e
u
si
n
g
a
h
y
b
r
i
d
Y
O
LO
a
n
d
R
e
sN
e
t
n
e
t
w
o
r
k
s
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
r
t
s
,
v
o
l
.
1
4
,
n
o
.
1
,
p
p
.
1
–
2
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s4
1
5
9
8
-
0
2
4
-
5
4
8
6
4
-
6.
[
1
2
]
J.
H
.
L
e
e
,
J.
W
.
C
h
a
e
,
a
n
d
H
.
C
.
C
h
o
,
“
I
mp
r
o
v
e
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
o
f
d
i
f
f
e
r
e
n
t
b
r
a
i
n
t
u
m
o
r
s
i
n
M
R
I
sc
a
n
s
u
si
n
g
P
a
t
t
e
r
n
e
d
-
G
r
i
d
M
a
s
k
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
4
0
2
0
4
–
4
0
2
1
2
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
3
7
7
1
0
5
.
[
1
3
]
J.
Ja
i
n
,
M
.
K
u
b
a
d
i
a
,
M
.
M
a
n
g
l
a
,
a
n
d
P
.
Ta
w
d
e
,
“
C
o
mp
a
r
i
s
o
n
o
f
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
t
o
c
l
a
ssi
f
y
b
r
a
i
n
t
u
mo
u
r
s
u
s
i
n
g
M
R
I
i
ma
g
e
s,
”
E
n
g
i
n
e
e
ri
n
g
Pr
o
c
e
e
d
i
n
g
s
,
v
o
l
.
5
9
,
n
o
.
1
,
p
p
.
1
–
9
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
e
n
g
p
r
o
c
2
0
2
3
0
5
9
1
4
4
.
[
1
4
]
B
.
S
a
r
a
d
a
,
K
.
N
.
R
e
d
d
y
,
R
.
B
a
b
u
,
a
n
d
B
.
S
.
R
.
B
a
b
u
,
“
B
r
a
i
n
t
u
m
o
r
c
l
a
s
si
f
i
c
a
t
i
o
n
u
si
n
g
mo
d
i
f
i
e
d
R
e
sN
e
t
5
0
V
2
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
m
p
u
t
i
n
g
a
n
d
D
i
g
i
t
a
l
S
y
st
e
m
s
,
v
o
l
.
1
6
,
n
o
.
1
,
p
p
.
1
–
1
0
,
2
0
2
4
.
[
1
5
]
M
.
G
ü
l
e
r
a
n
d
E.
N
a
m
l
ı
,
“
B
r
a
i
n
t
u
m
o
r
d
e
t
e
c
t
i
o
n
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
me
t
h
o
d
s’
c
l
a
s
si
f
i
e
r
o
p
t
i
m
i
z
a
t
i
o
n
u
si
n
g
med
i
c
a
l
i
ma
g
e
s
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
4
,
n
o
.
2
,
p
p
.
1
–
2
5
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
4
0
2
0
6
4
2
.
[
1
6
]
G
.
D
.
K
u
mar
a
n
d
S
.
N
.
M
o
h
a
n
t
y
,
“
B
r
a
i
n
t
u
mo
r
d
e
t
e
c
t
i
o
n
b
a
se
d
o
n
m
u
l
t
i
p
l
e
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
s fo
r
M
R
I
i
ma
g
e
s,”
EAI E
n
d
o
rs
e
d
T
ra
n
s
a
c
t
i
o
n
s
o
n
Pe
r
v
a
si
v
e
H
e
a
l
t
h
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
0
,
p
p
.
1
–
8
,
2
0
2
4
,
d
o
i
:
1
0
.
4
1
0
8
/
e
e
t
p
h
t
.
1
0
.
5
4
9
9
.
[
1
7
]
M
.
R
o
y
,
B
.
U
p
a
d
h
y
a
y
a
,
J.
R
a
i
,
a
n
d
K
.
S
h
a
r
m
a
,
“
R
e
s
o
l
u
t
i
o
n
e
n
h
a
n
c
e
m
e
n
t
o
f
b
r
a
i
n
M
R
I
i
ma
g
e
s u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
,
”
E
n
g
i
n
e
e
ri
n
g
Pro
c
e
e
d
i
n
g
s
,
v
o
l
.
5
9
,
n
o
.
1
,
p
p
.
1
–
8
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
e
n
g
p
r
o
c
2
0
2
3
0
5
9
1
5
8
.
[
1
8
]
S
.
K
.
M
a
t
h
i
v
a
n
a
n
,
S
.
S
o
n
a
i
m
u
t
h
u
,
S
.
M
u
r
u
g
e
sa
n
,
H
.
R
a
j
a
d
u
r
a
i
,
B
.
D
.
S
h
i
v
a
h
a
r
e
,
a
n
d
M
.
A
.
S
h
a
h
,
“
Em
p
l
o
y
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
f
o
r
a
c
c
u
r
a
t
e
b
r
a
i
n
t
u
mo
r
d
e
t
e
c
t
i
o
n
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
r
t
s
,
v
o
l
.
1
4
,
n
o
.
1
,
p
p
.
1
–
1
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s4
1
5
9
8
-
024
-
5
7
9
7
0
-
7.
[
1
9
]
S
.
A
l
-
O
t
a
i
b
i
,
A
.
R
e
h
m
a
n
,
A
.
R
a
z
a
,
J.
A
l
y
a
m
i
,
a
n
d
T.
S
a
b
a
,
“
C
V
G
-
N
e
t
:
n
o
v
e
l
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
b
a
se
d
d
e
e
p
f
e
a
t
u
r
e
s
f
o
r
d
i
a
g
n
o
si
s
o
f
b
r
a
i
n
t
u
m
o
r
s
u
si
n
g
M
R
I
sca
n
s,
”
P
e
e
rJ
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
0
,
p
p
.
1
–
2
4
,
2
0
2
4
,
d
o
i
:
1
0
.
7
7
1
7
/
p
e
e
r
j
-
c
s.
2
0
0
8
.
[
2
0
]
M
.
N
a
v
e
e
n
k
u
mar,
G
.
S
u
d
h
a
k
a
r
a
n
,
K
.
M
.
S
a
n
d
h
a
n
a
v
i
m
a
l
,
a
n
d
E
.
L
.
V
i
m
a
l
r
a
j
,
“
R
e
sN
e
t
mo
d
e
l
s
f
o
r
b
r
a
i
n
t
u
mo
r
se
g
me
n
t
a
t
i
o
n
,
”
J
o
u
rn
a
l
o
f
Ph
a
rm
a
c
e
u
t
i
c
a
l
N
e
g
a
t
i
v
e
Re
su
l
t
s
,
v
o
l
.
1
4
,
n
o
.
4
,
p
p
.
8
6
8
–
8
7
6
,
2
0
2
4
,
d
o
i
:
1
0
.
4
7
7
5
0
/
8
t
a
k
z
4
6
9
.
[
2
1
]
M
.
S
.
U
l
l
a
h
,
M
.
A
.
K
h
a
n
,
N
.
A
.
A
l
m
u
j
a
l
l
y
,
M
.
A
l
h
a
i
s
o
n
i
,
T.
A
k
r
a
m,
a
n
d
M
.
S
h
a
b
a
z
,
“
B
r
a
i
n
N
e
t
:
A
f
u
s
i
o
n
a
ssi
s
t
e
d
n
o
v
e
l
o
p
t
i
ma
l
f
r
a
mew
o
r
k
o
f
r
e
si
d
u
a
l
b
l
o
c
k
s
a
n
d
s
t
a
c
k
e
d
a
u
t
o
e
n
c
o
d
e
r
s
f
o
r
m
u
l
t
i
mo
d
a
l
b
r
a
i
n
t
u
m
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
rt
s
,
v
o
l
.
1
4
,
n
o
.
1
,
p
p
.
1
–
2
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
8
-
0
2
4
-
5
6
6
5
7
-
3.
[
2
2
]
Y
.
Y
a
n
g
,
P
.
W
a
n
g
,
Z.
Y
a
n
g
,
Y
.
Ze
n
g
,
F
.
C
h
e
n
,
Z
.
W
a
n
g
,
a
n
d
S
.
R
i
z
z
o
,
“
S
e
g
me
n
t
a
t
i
o
n
m
e
t
h
o
d
o
f
ma
g
n
e
t
i
c
r
e
so
n
a
n
c
e
i
ma
g
i
n
g
b
r
a
i
n
t
u
mo
r
i
ma
g
e
s
b
a
se
d
o
n
i
m
p
r
o
v
e
d
U
N
e
t
n
e
t
w
o
r
k
,
”
T
r
a
n
sl
a
t
i
o
n
a
l
C
a
n
c
e
r
Re
s
e
a
rc
h
,
v
o
l
.
1
3
,
n
o
.
3
,
p
p
.
1
5
6
7
–
1
5
8
3
,
2
0
2
4
,
d
o
i
:
1
0
.
2
1
0
3
7
/
t
c
r
-
23
-
1
8
5
8
.
[
2
3
]
N
.
Y
a
q
o
o
b
e
t
a
l
.
,
“
P
r
e
d
i
c
t
i
o
n
o
f
A
l
z
h
e
i
m
e
r
's
d
i
se
a
se
s
t
a
g
e
s
b
a
s
e
d
o
n
R
e
sN
et
-
se
l
f
-
a
t
t
e
n
t
i
o
n
a
r
c
h
i
t
e
c
t
u
r
e
w
i
t
h
B
a
y
e
si
a
n
o
p
t
i
m
i
z
a
t
i
o
n
a
n
d
b
e
st
f
e
a
t
u
r
e
s
s
e
l
e
c
t
i
o
n
,
”
Fr
o
n
t
i
e
rs
i
n
C
o
m
p
u
t
a
t
i
o
n
a
l
N
e
u
r
o
sci
e
n
c
e
,
v
o
l
.
1
8
,
p
p
.
1
–
1
6
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
8
9
/
f
n
c
o
m
.
2
0
2
4
.
1
3
9
3
8
4
9
.
[
2
4
]
M
.
J
o
sh
i
a
n
d
B
.
K
.
S
i
n
g
h
,
“
P
r
o
p
o
r
t
i
o
n
e
s
t
i
m
a
t
i
o
n
a
n
d
mu
l
t
i
-
c
l
a
ss
c
l
a
s
s
i
f
i
c
a
t
i
o
n
o
f
a
b
n
o
r
mal
b
r
a
i
n
c
e
l
l
s,
”
M
e
d
i
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
1
,
n
o
.
2
,
p
p
.
7
9
–
9
0
,
2
0
2
4
,
d
o
i
:
1
0
.
4
7
8
5
2
/
b
o
n
v
i
e
w
M
ED
I
N
4
2
0
2
1
6
8
5
.
[
2
5
]
M
.
A
.
R
a
h
ma
n
,
A
.
G
h
o
s
h
,
M
.
S
.
R
a
h
ma
n
,
a
n
d
M
.
A
.
S
i
d
d
i
q
u
e
,
“
A
c
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
si
s
a
n
d
v
i
s
u
a
l
i
z
a
b
l
e
o
f
M
R
I
i
ma
g
e
t
y
p
e
-
b
a
s
e
d
b
r
a
i
n
t
u
m
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
m
o
d
e
l
s,”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
E
l
e
c
t
r
i
c
a
l
En
g
i
n
e
e
ri
n
g
a
n
d
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
(
I
C
E
EI
C
T
)
,
p
p
.
7
–
1
2
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
EEI
C
T6
2
0
1
6
.
2
0
2
4
.
1
0
5
3
4
4
5
7
.
[
2
6
]
M
.
Z.
K
h
a
l
i
k
i
a
n
d
M
.
S
.
B
a
şars
l
a
n
,
“
B
r
a
i
n
t
u
m
o
r
d
e
t
e
c
t
i
o
n
f
r
o
m
i
ma
g
e
s
a
n
d
c
o
m
p
a
r
i
so
n
w
i
t
h
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
me
t
h
o
d
s
a
n
d
3
-
l
a
y
e
r
C
N
N
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
rt
s
,
v
o
l
.
1
4
,
n
o
.
1
,
p
p
.
1
–
1
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s4
1
5
9
8
-
0
2
4
-
5
2
8
2
3
-
9.
[
2
7
]
N
.
P
a
r
i
m
a
l
a
a
n
d
G
.
M
u
n
e
e
sw
a
r
i
,
“
M
o
d
i
f
i
e
d
R
e
sN
e
t
1
5
2
v
2
:
B
i
n
a
r
y
c
l
a
ss
i
f
i
c
a
t
i
o
n
a
n
d
h
y
b
r
i
d
s
e
g
me
n
t
a
t
i
o
n
o
f
b
r
a
i
n
st
r
o
k
e
u
si
n
g
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
-
b
a
se
d
a
p
p
r
o
a
c
h
,
”
P
o
l
i
sh
J
o
u
r
n
a
l
o
f
M
e
d
i
c
a
l
P
h
y
si
c
s
a
n
d
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
3
0
,
n
o
.
1
,
p
p
.
2
4
–
3
5
,
2
0
2
4
,
d
o
i
:
1
0
.
2
4
7
8
/
p
j
m
p
e
-
2
0
2
4
-
0
0
0
4
.
[
2
8
]
I
.
B
.
S
a
n
t
o
so
a
n
d
S
.
N
.
U
t
a
ma
,
“
A
n
e
w
v
o
t
i
n
g
o
f
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
f
o
r
b
r
a
i
n
t
u
m
o
r
d
e
t
e
c
t
i
o
n
b
a
se
d
o
n
M
R
I
i
ma
g
e
s,
”
I
n
t
e
r
n
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
st
e
m
s
,
v
o
l
.
1
7
,
n
o
.
1
,
p
p
.
2
1
2
–
2
2
7
,
2
0
2
4
,
d
o
i
:
1
0
.
2
2
2
6
6
/
i
j
i
e
s2
0
2
4
.
0
2
2
9
.
2
1
.
[
2
9
]
A
.
Y
o
u
n
i
s
e
t
a
l
.
,
“
A
b
n
o
r
m
a
l
b
r
a
i
n
t
u
mo
r
s c
l
a
s
si
f
i
c
a
t
i
o
n
u
si
n
g
R
e
sN
e
t
5
0
a
n
d
i
t
s c
o
m
p
r
e
h
e
n
s
i
v
e
e
v
a
l
u
a
t
i
o
n
,
”
I
E
EE
A
c
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
7
8
8
4
3
–
7
8
8
5
3
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
4
0
3
9
0
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.