I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
rm
a
t
ics a
nd
Co
m
m
un
ica
t
io
n T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
,
p
p
.
1
4
5
9
~1
4
6
8
I
SS
N:
2252
-
8
7
7
6
,
DOI
:
1
0
.
1
1
5
9
1
/iji
ct
.
v15
i
4
.
pp
1
4
5
9
-
1
4
6
8
1459
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
SSDH
R clas
sifie
r
network
for bra
in
t
umo
r d
etec
tion
using
M
RI ima
g
es
K
a
lpa
na
T
elk
a
r
,
K
.
Anus
ud
ha
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s E
n
g
i
n
e
e
r
i
n
g
,
P
o
n
d
i
c
h
e
r
r
y
U
n
i
v
e
r
si
t
y
,
P
o
n
d
i
c
h
e
r
r
y
,
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
Sep
3
,
2
0
2
5
R
ev
is
ed
J
u
n
1
,
2
0
2
6
Acc
ep
ted
J
u
l 5
,
2
0
2
6
Bra
in
tu
m
o
rs
a
re
th
e
c
h
a
o
t
ic
g
ro
wth
o
f
a
b
n
o
rm
a
l
b
ra
i
n
c
e
ll
s,
wh
ich
is
th
e
m
a
in
c
a
u
se
o
f
c
a
n
c
e
r
wo
rl
d
wid
e
.
S
y
m
p
to
m
s
v
a
ry
d
e
p
e
n
d
i
n
g
o
n
t
u
m
o
r
siz
e
a
n
d
lo
c
a
ti
o
n
,
a
n
d
th
e
se
tu
m
o
rs
c
a
n
b
e
c
a
n
c
e
ro
u
s
o
r
n
o
n
-
c
a
n
c
e
ro
u
s.
Earl
y
d
e
tec
ti
o
n
a
n
d
c
h
a
ra
c
teriz
a
ti
o
n
a
re
c
ru
c
i
a
l
b
u
t
c
h
a
ll
e
n
g
in
g
d
u
e
t
o
th
e
b
ra
i
n
'
s
c
o
m
p
lex
str
u
c
tu
re
.
T
h
is
st
u
d
y
p
r
o
p
o
se
s
a
se
lf
-
su
p
e
rv
ise
d
c
las
sific
a
ti
o
n
d
e
e
p
h
iera
rc
h
ica
l
re
c
o
n
stru
c
ti
o
n
n
e
two
rk
(S
S
DH
R),
a
c
o
m
b
i
n
a
ti
o
n
o
f
g
e
n
e
ra
ti
v
e
a
n
d
d
isc
rimin
a
ti
v
e
m
o
d
e
ls.
Th
i
s
n
e
two
rk
u
se
s
c
o
n
v
o
l
u
ti
o
n
a
l
K
e
rn
e
ls
o
f
d
iffere
n
t
siz
e
s
f
o
r
p
ro
c
e
ss
in
g
o
f
in
p
u
t
d
a
ta
o
n
se
p
a
ra
te
b
ra
n
c
h
e
s
to
e
x
trac
t
e
ffe
c
ti
v
e
d
isc
rimin
a
ti
v
e
fe
a
tu
re
s.
A
p
a
ra
ll
e
l
-
stru
c
tu
re
d
sy
m
m
e
tr
ic
fu
sio
n
a
tt
e
n
ti
o
n
(S
F
A)
m
o
d
u
le
is
in
tro
d
u
c
e
d
,
c
o
m
b
in
in
g
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
s
fo
r
c
o
o
rd
i
n
a
tes
,
ti
m
e
,
a
n
d
fre
q
u
e
n
c
y
to
e
n
h
a
n
c
e
th
e
m
o
d
e
l'
s
c
o
m
p
re
h
e
n
sio
n
a
n
d
fe
a
tu
re
se
lec
ti
o
n
.
T
h
e
S
S
DH
R
c
las
sifier
n
e
two
rk
a
c
h
iev
e
s
t
h
e
h
i
g
h
e
st
a
c
c
u
ra
c
y
(9
1
.
9
%
)
a
m
o
n
g
1
0
sta
te
-
of
-
th
e
-
a
rt
c
las
s
ifi
e
rs,
c
las
sify
i
n
g
t
u
m
o
rs
in
to
4
c
las
se
s
:
p
it
u
it
a
ry
,
m
e
n
i
n
g
io
m
a
,
g
li
o
m
a
,
a
n
d
n
o
tu
m
o
r
.
Th
e
m
o
d
e
l'
s
p
e
rfo
rm
a
n
c
e
is
e
v
a
lu
a
ted
u
si
n
g
v
a
ri
o
u
s
m
e
tri
c
s
th
a
t
d
e
m
o
n
stra
te
it
s
e
ffe
c
ti
v
e
n
e
ss
in
b
ra
i
n
tu
m
o
r
c
las
sifica
ti
o
n
.
K
ey
w
o
r
d
s
:
B
r
ain
tu
m
o
r
s
C
las
s
if
icatio
n
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
Me
tr
ics
Ma
g
n
etic
r
eso
n
an
ce
im
a
g
in
g
im
ag
es
SS
DH
R
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
:
Kalp
an
a
T
elk
ar
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics E
n
g
i
n
ee
r
in
g
,
Po
n
d
ich
e
r
r
y
U
n
iv
er
s
ity
Po
n
d
ich
er
r
y
,
I
n
d
ia
E
m
ail: te
lk
ar
k
alp
an
a@
g
m
ail.
c
o
m
1.
I
NT
RO
D
UCT
I
O
N
B
r
ain
tu
m
o
r
class
if
icatio
n
is
i
m
p
o
r
tan
t
f
o
r
id
e
n
tify
in
g
tu
m
o
r
ty
p
es
s
u
c
h
as
Pit
u
itar
y
,
m
e
n
in
g
io
m
a,
an
d
g
lio
m
a,
an
d
f
o
r
s
elec
tin
g
s
u
itab
le
tr
ea
tm
e
n
ts
.
Am
o
n
g
v
ar
io
u
s
im
a
g
in
g
tech
n
i
q
u
es,
m
ag
n
etic
r
eso
n
an
c
e
im
ag
in
g
(
MRI
)
is
wid
ely
p
r
ef
er
r
ed
b
ec
au
s
e
it
p
r
o
v
id
es
h
ig
h
-
q
u
ality
im
a
g
es
with
o
u
t
u
s
in
g
io
n
izin
g
r
a
d
iatio
n
[
1
]
.
A
b
r
ain
tu
m
o
r
is
an
u
n
c
o
n
tr
o
lled
g
r
o
wth
o
f
b
r
ain
ce
lls
an
d
is
co
n
s
id
er
ed
o
n
e
o
f
t
h
e
m
o
s
t
life
-
th
r
ea
ten
in
g
f
o
r
m
s
o
f
ca
n
ce
r
[
2
]
.
Acc
o
r
d
i
n
g
to
c
an
ce
r
s
tatis
tics
in
th
e
USA,
ar
o
u
n
d
2
3
,
0
0
0
b
r
ain
tu
m
o
r
c
ases
wer
e
r
ep
o
r
ted
in
2
0
1
5
,
m
ak
i
n
g
it
a
m
aj
o
r
ca
u
s
e
o
f
ca
n
ce
r
-
r
elate
d
d
ea
th
s
i
n
b
o
th
ch
ild
r
en
an
d
a
d
u
lts
[
3
]
.
R
ad
io
lo
g
is
ts
u
s
e
s
ev
er
al
m
ed
ical
im
ag
in
g
m
eth
o
d
s
f
o
r
d
ia
g
n
o
s
is
,
with
MRI
b
ein
g
th
e
m
o
s
t
ef
f
ec
tiv
e
an
d
s
af
est
o
p
tio
n
[
4
]
.
T
r
ad
itio
n
a
l
tu
m
o
r
class
if
icatio
n
is
tim
e
-
co
n
s
u
m
in
g
,
co
s
tly
,
an
d
h
i
g
h
ly
d
ep
en
d
en
t
o
n
r
ad
io
lo
g
is
ts
,
wh
ich
ca
n
lead
to
d
iag
n
o
s
tic
er
r
o
r
s
wh
en
h
an
d
lin
g
lar
g
e
v
o
lu
m
es
o
f
MRI
d
ata.
T
o
o
v
er
co
m
e
th
ese
lim
itatio
n
s
,
r
esear
ch
er
s
ar
e
d
ev
elo
p
in
g
au
to
m
ated
i
m
ag
e
p
r
o
ce
s
s
in
g
an
d
co
m
p
u
ter
-
aid
ed
d
i
ag
n
o
s
is
(
C
AD)
s
y
s
tem
s
th
at
r
ed
u
ce
h
u
m
an
er
r
o
r
a
n
d
wo
r
k
lo
a
d
[
5
]
–
[
8
]
.
Stu
d
ies
s
h
o
w
th
at
d
ee
p
lear
n
in
g
(
DL
)
tech
n
iq
u
es
im
p
r
o
v
e
t
h
e
p
er
f
o
r
m
an
ce
o
f
C
AD
s
y
s
tem
s
[
9
]
.
Ho
wev
er
,
lim
ited
m
e
d
ical
d
at
asets
m
ak
e
tr
ain
in
g
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
et
wo
r
k
(
C
NN
)
m
o
d
els f
r
o
m
s
cr
a
tch
d
if
f
icu
lt a
n
d
tim
e
-
co
n
s
u
m
i
n
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J
I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
4
5
9
-
1
4
6
8
1460
2.
RELA
TED WO
RK
C
NNs
h
av
e
s
h
o
wn
s
tr
o
n
g
p
er
f
o
r
m
an
ce
in
b
r
ai
n
tu
m
o
r
d
ete
ctio
n
an
d
class
if
icatio
n
u
s
in
g
MRI
s
ca
n
s
d
u
e
to
ad
v
an
ce
s
in
DL
.
Pas
h
ar
i
et
a
l.
[
1
0
]
co
m
b
in
ed
C
NN
f
ea
tu
r
es
with
Ker
n
el
ex
tr
e
m
e
lear
n
in
g
m
ac
h
in
e
(
KE
L
M)
an
d
ac
h
ie
v
ed
9
3
.
6
8
%
ac
cu
r
ac
y
o
n
3
0
6
4
MRI
i
m
ag
es.
Das
et
a
l.
[
1
1
]
,
Gau
s
s
ian
f
ilter
in
g
an
d
h
is
to
g
r
am
eq
u
aliza
tio
n
wer
e
ap
p
lied
b
ef
o
r
e
C
NN
clas
s
if
icatio
n
,
r
esu
ltin
g
in
9
4
.
3
9
%
ac
cu
r
ac
y
.
T
r
a
n
s
f
er
lear
n
in
g
with
Go
o
g
leNe
t
an
d
SVM
in
[
1
2
]
ac
h
ie
v
ed
9
7
.
1
%
ac
cu
r
ac
y
o
n
th
e
Fig
Sh
ar
e
d
ata
s
et.
Sev
er
al
s
tu
d
ies
p
r
o
p
o
s
ed
ad
v
an
ce
d
C
NN
ar
ch
itectu
r
es
f
o
r
class
if
y
in
g
p
itu
itar
y
,
g
lio
m
a,
an
d
m
en
in
g
io
m
a
tu
m
o
r
s
.
A
C
NN
-
b
ased
m
o
d
el
in
[
1
3
]
ac
h
iev
e
d
9
6
.
5
6
%
ac
cu
r
ac
y
,
wh
ile
a
d
if
f
er
en
tial
d
ee
p
C
NN
in
[
1
4
]
r
ea
ch
ed
9
9
.
2
5
%
ac
cu
r
ac
y
o
n
2
5
,
0
0
0
MRI
im
ag
es.
An
o
th
er
C
NN
m
o
d
el
in
[
1
5
]
ac
h
ie
v
ed
9
9
.
4
%
class
if
icati
o
n
ac
cu
r
ac
y
.
Sajjad
et
a
l.
[
1
6
]
,
C
NN
-
b
ased
s
eg
m
en
tatio
n
with
d
ata
au
g
m
en
tat
io
n
an
d
VGG
-
1
9
f
in
e
-
t
u
n
in
g
to
im
p
r
o
v
e
m
u
lti
-
g
r
ad
e
tu
m
o
r
class
if
icatio
n
.
Hy
b
r
id
an
d
tr
an
s
f
er
lear
n
in
g
ap
p
r
o
ac
h
es
also
p
r
o
d
u
ce
d
s
tr
o
n
g
r
esu
lts
.
C
NN
co
m
b
in
ed
with
L
STM
i
n
[
1
7
]
ac
h
iev
ed
9
7
%
ac
c
u
r
ac
y
o
n
a
Kag
g
le
d
ataset.
Dee
p
C
NN
ar
ch
itectu
r
es
with
m
u
ltip
le
co
n
v
o
lu
tio
n
a
n
d
p
o
o
l
in
g
lay
e
r
s
wer
e
ex
p
lo
r
ed
in
[
1
8
]
,
w
h
ile
[
1
9
]
co
m
p
ar
ed
th
r
ee
C
NN
m
o
d
els
with
ac
cu
r
ac
ies
u
p
to
9
9
.
3
3
%.
T
r
an
s
f
er
lear
n
in
g
m
o
d
els
s
u
ch
as
R
esNe
t5
0
,
VGG1
9
,
an
d
I
n
ce
p
tio
n
-
v
3
in
[
2
0
]
ac
h
iev
ed
ar
o
u
n
d
9
0
%
ac
c
u
r
ac
y
.
Oth
er
ap
p
r
o
ac
h
es
in
clu
d
e
C
NN
clas
s
if
icatio
n
o
f
th
r
ee
t
u
m
o
r
ty
p
es
in
[
2
1
]
,
ac
h
iev
in
g
9
8
.
5
1
%
tr
ai
n
in
g
ac
cu
r
ac
y
,
an
d
m
ac
h
in
e
lea
r
n
in
g
-
b
ased
class
if
icatio
n
in
[
2
2
]
with
9
1
.
4
3
%
cr
o
s
s
-
v
alid
atio
n
ac
cu
r
ac
y
.
C
o
n
ten
t
-
b
ased
im
ag
e
r
etr
ie
v
al
tech
n
iq
u
es
u
s
in
g
MI
D
an
d
MPP
alg
o
r
ith
m
s
in
[
2
3
]
ac
h
iev
ed
h
ig
h
p
r
ec
is
io
n
o
n
C
E
-
MRI
d
atasets
.
B
eh
in
et
a
l.
[
2
4
]
an
d
Dr
ev
ele
g
as
[
2
5
]
f
o
c
u
s
ed
o
n
g
lio
m
a
g
en
etics
an
d
tu
m
o
r
e
v
alu
atio
n
m
eth
o
d
s
,
wh
ile
[
2
6
]
,
[
2
7
]
d
is
cu
s
s
ed
d
ee
p
C
NN
ar
ch
i
tectu
r
es
f
o
r
im
ag
e
class
if
icatio
n
,
d
etec
tio
n
,
an
d
s
eg
m
en
tatio
n
.
3.
ST
A
T
E
-
OF
-
T
H
E
-
ART CN
N
M
O
DE
L
S
I
n
th
is
w
o
r
k
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
is
co
m
p
ar
ed
with
n
in
e
e
x
i
s
tin
g
C
NN
-
b
ased
ar
ch
itectu
r
e
s
u
s
in
g
th
e
s
am
e
Kag
g
le
d
ataset.
−
I
n
ce
p
tio
n
-
R
esNet
-
v
2
co
m
b
in
es
I
n
ce
p
tio
n
m
o
d
u
les
with
r
esid
u
al
co
n
n
ec
tio
n
s
to
im
p
r
o
v
e
m
u
lti
-
s
ca
le
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
o
v
e
r
co
m
e
v
an
is
h
in
g
g
r
ad
ie
n
t iss
u
es i
n
d
ee
p
n
etwo
r
k
s
[
2
8
]
.
−
L
eNe
t
-
5
,
in
tr
o
d
u
ce
d
b
y
Yan
n
L
eCu
n
in
1
9
9
8
,
is
an
ea
r
ly
C
NN
ar
ch
itectu
r
e
co
n
s
is
tin
g
o
f
co
n
v
o
lu
tio
n
,
p
o
o
lin
g
,
an
d
f
u
lly
c
o
n
n
ec
ted
lay
er
s
,
wid
ely
u
s
ed
f
o
r
u
n
d
er
s
t
an
d
in
g
C
NN
f
u
n
d
am
en
tals
[
2
9
]
.
−
R
esNet1
0
1
-
L
STM
in
teg
r
ates
R
esNet1
0
1
f
o
r
s
p
atial
f
ea
tu
r
e
ex
tr
ac
tio
n
with
L
STM
f
o
r
le
ar
n
in
g
tem
p
o
r
al
p
atter
n
s
in
s
eq
u
en
tial M
R
I
d
at
a
[
3
0
]
.
−
I
n
ce
p
t
io
n
v
3
im
p
r
o
v
es
class
i
f
icatio
n
p
er
f
o
r
m
an
ce
th
r
o
u
g
h
f
ac
to
r
ized
co
n
v
o
lu
tio
n
s
,
r
eg
u
lar
izatio
n
,
an
d
au
x
iliar
y
class
if
ier
s
[
3
1
]
.
−
Xce
p
tio
n
em
p
l
o
y
s
d
e
p
th
-
wis
e
s
ep
ar
ab
le
c
o
n
v
o
lu
tio
n
s
,
im
p
r
o
v
in
g
n
etwo
r
k
ef
f
icien
cy
a
n
d
p
e
r
f
o
r
m
an
ce
[
3
2
]
.
−
R
esNet1
0
1
-
L
STM
an
d
GR
U
co
m
b
in
es
L
STM
an
d
GR
U
u
n
its
to
ca
p
tu
r
e
lo
n
g
-
ter
m
d
ep
en
d
en
cies
an
d
tem
p
o
r
al
f
ea
tu
r
es e
f
f
icien
tly
i
n
m
ed
ical
d
ata
[
3
3
]
.
−
C
o
n
v
NeX
t
B
ase
m
o
d
er
n
izes
C
NN
ar
ch
itectu
r
e
u
s
in
g
tr
an
s
f
o
r
m
er
-
in
s
p
ir
ed
co
n
ce
p
ts
s
u
ch
as
d
ep
th
-
wis
e
co
n
v
o
l
u
tio
n
an
d
lay
er
n
o
r
m
aliza
tio
n
f
o
r
ef
f
e
ctiv
e
m
e
d
ical
im
ag
e
class
if
icatio
n
[
3
4
]
.
−
VGG1
9
,
d
e
v
elo
p
ed
b
y
th
e
Vis
u
al
Geo
m
etr
y
Gr
o
u
p
at
th
e
Un
iv
er
s
ity
o
f
Ox
f
o
r
d
,
u
s
es
d
ee
p
3
×3
co
n
v
o
l
u
tio
n
lay
er
s
f
o
r
ac
c
u
r
at
e
im
ag
e
f
ea
tu
r
e
lear
n
in
g
[
3
5
]
.
−
R
esNet1
0
1
is
a
d
ee
p
r
esid
u
al
n
etwo
r
k
with
1
0
1
lay
er
s
th
a
t
lear
n
s
h
ig
h
-
lev
el
MRI
f
ea
tu
r
es
ef
f
ec
tiv
ely
u
s
in
g
r
esid
u
al
co
n
n
ec
tio
n
s
[
3
6
]
.
T
ab
le
1
s
u
m
m
ar
izes
th
e
tr
ain
a
b
le
an
d
n
o
n
-
tr
ain
a
b
le
p
ar
a
m
et
er
s
,
to
tal
p
ar
a
m
eter
s
,
an
d
m
e
m
o
r
y
u
s
ag
e
o
f
ea
c
h
m
o
d
el.
T
r
ain
ab
le
p
ar
a
m
eter
s
ar
e
u
p
d
ated
d
u
r
in
g
tr
ain
in
g
,
wh
er
ea
s
n
o
n
-
tr
ain
ab
le
p
ar
a
m
eter
s
r
em
ain
f
ix
ed
.
T
h
e
tab
le
also
h
ig
h
lig
h
ts
th
e
c
o
m
p
lex
ity
,
s
tr
en
g
th
s
,
wea
k
n
es
s
es,
an
d
n
o
v
elty
o
f
ea
c
h
ar
ch
it
ec
tu
r
e.
4.
P
RO
P
O
SE
D
M
E
T
H
O
DO
L
O
G
Y
T
h
is
s
ec
tio
n
d
escr
ib
es
in
d
etai
l
th
e
s
u
g
g
ested
a
p
p
r
o
ac
h
.
T
h
e
r
ich
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
ar
e
o
b
tain
ed
b
y
co
m
b
in
i
n
g
d
is
cr
im
in
ativ
e
an
d
g
en
er
ativ
e
tech
n
iq
u
e
s
.
T
o
im
p
r
o
v
e
th
e
d
is
cr
im
in
ab
ilit
y
b
etwe
en
ch
ar
ac
ter
is
tics
o
f
n
o
r
m
al
a
n
d
ab
n
o
r
m
al
im
a
g
es,
m
o
d
u
les
f
o
r
atten
tio
n
an
d
m
i
x
ed
ce
n
ter
lo
s
s
ar
e
p
r
o
p
o
s
ed
.
C
o
n
ca
ten
ated
f
ea
tu
r
es a
r
e
th
e
last
m
eth
o
d
u
s
ed
f
o
r
b
r
ain
tu
m
o
r
d
etec
tio
n
.
4
.
1
.
Self
-
s
up
er
v
is
ed
cla
s
s
if
ica
t
io
n
us
ing
deep
hiera
rc
hica
l
re
co
ns
t
ruct
io
n
net
wo
rk
wit
h
s
y
m
m
et
ric
f
us
io
n a
t
t
ent
io
n
Fig
u
r
e
1
s
h
o
ws
th
e
SS
DHR
c
lass
if
ier
n
etwo
r
k
s
tr
u
ctu
r
e
u
s
ed
f
o
r
b
r
ain
tu
m
o
r
d
etec
tio
n
.
T
h
e
m
o
d
el
u
s
es
b
o
th
class
i
f
icatio
n
f
ea
t
u
r
es
(
y
)
an
d
r
ec
o
n
s
tr
u
cted
l
aten
t
r
ep
r
esen
tatio
n
s
(
z)
to
im
p
r
o
v
e
clo
s
ed
-
s
et
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
S
S
DHR
cla
s
s
ifier
n
etw
o
r
k
fo
r
b
r
a
in
tu
mo
r
d
etec
tio
n
u
s
in
g
MRI
ima
g
es
(
K
a
lp
a
n
a
Telka
r
)
1461
class
if
icatio
n
an
d
tu
m
o
r
id
en
t
if
icatio
n
[
2
6
]
.
I
t
ex
tr
a
cts
late
n
t
f
ea
tu
r
es
f
r
o
m
m
id
d
le
lay
e
r
s
o
f
th
e
n
etwo
r
k
to
r
ed
u
ce
in
f
o
r
m
ati
o
n
lo
s
s
th
at
u
s
u
ally
o
cc
u
r
s
in
h
i
g
h
er
la
y
e
r
s
.
T
h
e
n
etwo
r
k
c
o
n
tain
s
two
p
ar
allel
s
y
m
m
etr
ic
b
r
an
ch
es
with
d
if
f
er
e
n
t
co
n
v
o
lu
tio
n
K
er
n
el
s
izes
(
3
×3
an
d
7
×7
)
,
allo
win
g
th
e
m
o
d
el
to
ca
p
tu
r
e
f
ea
tu
r
es
at
m
u
ltip
le
s
ca
les
an
d
im
p
r
o
v
e
r
o
b
u
s
tn
ess
an
d
f
ea
tu
r
e
r
ep
r
es
en
tatio
n
.
T
h
e
f
ea
tu
r
es
z1
to
z
6
r
ep
r
esen
t
p
o
o
led
o
u
tp
u
ts
f
r
o
m
r
ec
o
n
s
tr
u
cte
d
r
ep
r
esen
tatio
n
s
,
wh
ile
“Atten
tio
n
”
r
ef
er
s
to
th
e
p
r
o
p
o
s
ed
s
y
m
m
etr
ic
f
u
s
io
n
atten
tio
n
(
SF
A)
m
o
d
u
le.
Feat
u
r
es
f
r
o
m
b
o
th
b
r
an
ch
es
ar
e
c
o
m
b
in
ed
al
o
n
g
th
e
c
h
an
n
el
d
im
en
s
io
n
an
d
p
ass
ed
th
r
o
u
g
h
two
f
u
lly
c
o
n
n
ec
te
d
l
ay
er
s
f
o
r
class
if
icatio
n
.
T
h
e
S
FA
m
o
d
u
le
h
elp
s
th
e
m
o
d
el
f
o
cu
s
o
n
im
p
o
r
tan
t
in
p
u
t
f
ea
tu
r
es
an
d
im
p
r
o
v
es
o
v
er
all
p
er
f
o
r
m
an
ce
,
wh
ile
m
i
x
ed
ce
n
ter
l
o
s
s
is
u
s
ed
to
g
en
er
ate
m
o
r
e
c
o
m
p
ac
t
an
d
d
is
cr
im
in
ativ
e
f
e
atu
r
es
,
as sh
o
wn
in
Fig
u
r
e
2
.
Fig
u
r
e
1
.
T
h
e
ar
ch
itectu
r
e
o
f
S
SDHR
cla
s
s
if
ier
n
etwo
r
k
Fig
u
r
e
2
.
SF
A
m
o
d
el
a
r
ch
itect
u
r
e.
"G"
s
tan
d
s
f
o
r
th
e
s
p
lit g
r
o
u
p
in
g
s
4
.
2
.
Sy
m
m
e
t
ric
f
us
io
n a
t
t
ent
io
n
Fo
r
b
etter
d
etec
tio
n
o
f
b
r
ain
t
u
m
o
r
f
ea
tu
r
es,
th
e
p
r
o
p
o
s
ed
wo
r
k
r
elies
o
n
SF
A.
Fig
u
r
e
2
d
ep
icts
th
e
g
en
er
al
lay
o
u
t
o
f
th
e
atten
ti
o
n
m
o
d
u
le,
wh
ich
is
s
u
itab
le
f
o
r
p
ar
allel
-
s
tr
u
ctu
r
e
d
n
et
wo
r
k
s
.
T
o
p
r
eser
v
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J
I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
4
5
9
-
1
4
6
8
1462
in
f
o
r
m
atio
n
th
r
o
u
g
h
o
u
t
ea
ch
c
h
an
n
el
a
n
d
lo
we
r
co
m
p
u
tatio
n
al
o
v
er
h
e
ad
,
r
estru
ctu
r
e
a
p
o
r
t
io
n
o
f
th
e
c
h
an
n
els
an
d
ea
ch
in
p
u
t
f
ea
tu
r
e
X
∈
ℝ
C
xHx
W
in
to
b
atch
d
im
en
s
io
n
s
[
2
7
]
.
I
n
s
tead
o
f
ad
d
i
n
g
m
o
r
e
b
r
an
c
h
in
g
s
tr
u
ctu
r
es,
u
s
e
th
e
n
etwo
r
k
'
s
in
itial
p
ar
al
lel
b
r
an
c
h
es
to
ca
p
tu
r
e
m
u
ltis
ca
le
in
f
o
r
m
atio
n
.
Ad
d
itio
n
all
y
,
g
i
v
e
ea
c
h
b
r
an
ch
s
im
u
ltan
eo
u
s
atten
tio
n
to
b
o
th
f
r
eq
u
e
n
cy
a
n
d
tem
p
o
r
al
asp
ec
ts
.
Fig
u
r
e
3
s
h
o
ws
h
o
w
th
ese
two
atten
tio
n
s
ar
e
co
m
p
u
ted
.
C
o
n
s
eq
u
en
tly
,
ea
c
h
b
r
a
n
ch
o
f
t
h
e
s
y
m
m
etr
ical
s
tr
u
ctu
r
e
s
im
u
ltan
eo
u
s
ly
in
co
r
p
o
r
ates
th
r
ee
s
u
b
-
b
r
an
c
h
es,
n
am
ely
co
o
r
d
in
ate
atten
tio
n
,
tem
p
o
r
al
atten
tio
n
,
an
d
f
r
eq
u
en
cy
atten
tio
n
.
C
o
o
r
d
in
ate
atten
ti
o
n
(
C
A)
[
3
0
]
,
in
wh
ich
p
r
o
ce
s
s
in
g
o
f
in
p
u
ts
is
d
o
n
e
v
ia
one
-
d
im
e
n
s
io
n
al
h
o
r
izo
n
tal
a
n
d
v
er
tical
g
lo
b
al
p
o
o
lin
g
r
at
h
er
th
an
g
lo
b
al
a
v
er
ag
e
p
o
o
lin
g
.
T
h
is
allo
ws
th
e
p
r
eser
v
atio
n
o
f
ex
ac
t
p
o
s
itio
n
al
in
f
o
r
m
atio
n
alo
n
g
o
n
e
s
p
atial
g
u
id
an
ce
wh
ile
co
llectin
g
lo
n
g
-
r
a
n
g
e
d
ep
en
d
e
n
cies a
lo
n
g
a
n
o
th
er
.
At
h
eig
h
t h
,
t
h
e
c
-
th
c
h
an
n
el'
s
o
u
tp
u
t is g
iv
e
n
as:
ℎ
(
ℎ
)
=
1
∑
0
≤
≤
(
ℎ
,
)
(
1
)
T
h
e
c
-
th
c
h
an
n
el
o
u
tp
u
t w
ith
wid
th
w
ca
n
b
e
e
x
p
r
ess
ed
as:
(
)
=
1
∑
0
≤
≤
(
,
)
(
2
)
W
h
er
e
y
c
r
ep
r
esen
ts
th
e
c
-
th
c
h
an
n
el
in
p
u
t f
ea
tu
r
es.
Po
o
lin
g
th
e
r
esh
ap
ed
in
p
u
t
f
ea
tu
r
es
in
to
two
d
ir
ec
tio
n
s
,
th
en
p
er
f
o
r
m
i
n
g
1
x
1
co
n
v
o
lu
tio
n
al
p
r
o
ce
s
s
es
an
d
a
p
p
ly
in
g
t
h
e
a
ctiv
atio
n
f
u
n
ctio
n
as
s
ig
m
o
id
.
T
h
is
p
r
o
d
u
ce
s
CA
weig
h
ts
o
f
two
s
ets
ℎ
ℝ
⁄
×
×
1
an
d
ℝ
⁄
×
×
1
in
d
is
tin
ct
d
ir
ec
tio
n
s
.
C
A
ca
n
b
e
th
o
u
g
h
t
o
f
as
in
d
ep
en
d
en
tly
i
n
teg
r
atin
g
f
r
eq
u
e
n
cy
an
d
tim
e
in
f
o
r
m
a
tio
n
in
to
th
e
ch
an
n
el
d
im
e
n
s
io
n
wh
en
th
e
in
p
u
t
o
f
t
h
e
m
o
d
el
is
im
ag
e.
Fu
r
th
er
m
o
r
e
,
th
e
f
o
llo
win
g
ar
e
th
e
f
o
r
m
u
las f
o
r
t
h
e
p
r
e
v
io
u
s
ly
in
tr
o
d
u
ce
d
f
r
e
q
u
en
c
y
a
tten
tio
n
.
F
a
∈
ℝ
1
×
×
1
an
d
tem
p
o
r
al
atte
n
tio
n
is
T
a
∈
ℝ
1
×
×
1
:
=
(
(
(
1
,
1
)
(
2
1
1
(
)
)
1
,
3
)
)
1
,
3
(
3
)
=
(
(
(
1
,
1
)
(
2
1
1
(
)
)
1
,
2
)
)
1
,
2
(
4
)
wh
er
e
Av
g
Po
o
l
(
1
,
1
)
s
h
o
ws
t
h
e
ad
a
p
tiv
e
av
e
r
ag
e
p
o
o
lin
g
t
h
at
r
ed
u
ce
s
ea
ch
ch
an
n
el'
s
f
ea
tu
r
e
m
ap
to
1
x
1
,
(
σ)
d
escr
ib
es
th
e
s
ig
m
o
id
ac
tiv
ati
o
n
f
u
n
ctio
n
,
(
.
)
T1,
2
an
d
(
.
)
T1,
3
e
x
p
r
ess
es
th
e
tr
an
s
p
o
s
e
o
p
er
ati
o
n
s
o
n
d
im
e
n
s
io
n
s
1
an
d
2
a
n
d
d
i
m
en
s
io
n
s
1
an
d
3
,
r
esp
ec
tiv
ely
.
X
G
ℝ
⁄
×
×
d
escr
ib
es th
e
r
esh
ap
ed
in
p
u
t f
ea
tu
r
e
m
ap
.
Fig
u
r
e
3.
T
em
p
o
r
al
atten
tio
n
a
n
d
f
r
eq
u
en
cy
atten
tio
n
B
y
ap
p
ly
in
g
c
o
o
r
d
in
ate
atten
t
io
n
,
tem
p
o
r
al
atten
tio
n
,
an
d
f
r
eq
u
en
cy
atten
tio
n
to
X
G
an
d
weig
h
tin
g
th
e
m
,
t
h
e
o
u
tp
u
t th
at
is
o
b
tain
ed
is
X’
G.
ℝ
⁄
×
×
is
g
iv
en
as:
X
’
G
=
ℎ
+
X
G
T
a
+
G
F
a
(
5
)
+
+
=1
(
6
)
W
h
er
e
(
)
,
(
)
,
an
d
(
)
ar
e
lear
n
a
b
le
p
ar
am
eter
s
n
o
r
m
alize
d
u
s
in
g
th
e
So
f
tMa
x
f
u
n
ctio
n
.
I
n
f
o
r
m
atio
n
is
th
en
ag
g
r
eg
ated
ac
r
o
s
s
m
u
ltip
le
s
p
atial
d
im
en
s
io
n
s
i
n
a
cr
o
s
s
-
s
p
atial
m
an
n
er
.
T
h
e
two
s
y
m
m
etr
ical
b
r
a
n
ch
es
s
im
u
ltan
eo
u
s
ly
ap
p
ly
c
o
o
r
d
in
atio
n
,
tem
p
o
r
al,
an
d
f
r
eq
u
e
n
c
y
atten
tio
n
m
ec
h
a
n
is
m
s
.
Nex
t,
th
e
o
u
tp
u
ts
o
f
b
o
th
b
r
an
ch
es
u
n
d
er
g
o
two
-
d
im
en
s
io
n
al
g
lo
b
al
av
er
ag
e
p
o
o
lin
g
an
d
ar
e
r
esh
ap
ed
in
to
(
R
^{(
1
\
tim
es
C
/G)
})
.
T
h
e
b
r
an
ch
o
u
tp
u
ts
ar
e
f
u
r
th
er
r
e
s
h
ap
ed
in
to
(
R
^{(
C
/(
G
\
tim
es
HW
)
)
})
[
3
1
]
.
Sp
atial
atten
tio
n
m
ap
s
f
o
r
ea
c
h
b
r
an
ch
ar
e
g
en
er
ated
u
s
in
g
g
l
o
b
al
av
e
r
ag
e
p
o
o
lin
g
,
So
f
tMa
x
,
an
d
cr
o
s
s
-
b
r
an
ch
m
atr
ix
d
o
t
p
r
o
d
u
ct
o
p
er
atio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
S
S
DHR
cla
s
s
ifier
n
etw
o
r
k
fo
r
b
r
a
in
tu
mo
r
d
etec
tio
n
u
s
in
g
MRI
ima
g
es
(
K
a
lp
a
n
a
Telka
r
)
1463
T
h
e
s
p
atial
atten
tio
n
weig
h
ts
ar
e
o
b
tai
n
e
d
b
y
c
o
m
b
in
in
g
th
e
two
atten
tio
n
m
ap
s
an
d
a
p
p
ly
in
g
th
e
s
ig
m
o
id
f
u
n
ctio
n
.
Fin
ally
,
th
ese
weig
h
ts
ar
e
m
u
ltip
lied
with
th
e
co
r
r
esp
o
n
d
in
g
b
r
an
ch
i
n
p
u
ts
to
g
en
er
ate
o
u
tp
u
ts
h
av
in
g
t
h
e
s
am
e
d
im
en
s
io
n
s
a
s
th
e
in
p
u
ts
.
4
.
3
.
O
ne
cl
a
s
s
ce
ntr
e
lo
s
s
T
h
e
m
ain
g
o
al
o
f
ce
n
tr
e
lo
s
s
is
to
in
cr
ea
s
e
th
e
d
is
cr
im
in
ativ
e
s
tr
en
g
th
o
f
f
ea
tu
r
es
b
y
in
c
r
e
asin
g
th
e
d
is
tan
ce
b
etwe
en
f
ea
tu
r
e
p
o
in
ts
o
f
v
ar
io
u
s
class
es
an
d
b
y
d
ec
r
ea
s
in
g
th
e
d
is
tan
ce
b
etwe
en
th
e
f
ea
tu
r
e
p
o
in
ts
o
f
th
e
s
am
e
class
in
th
e
f
ea
tu
r
e
s
p
ac
e.
Ma
in
tain
in
g
a
ce
n
tr
e
f
o
r
ea
c
h
class
an
d
tr
ain
in
g
th
e
m
o
d
el
b
y
m
in
im
izin
g
t
h
e
s
ep
ar
atio
n
b
etwe
en
ea
ch
o
f
th
e
s
am
p
les
an
d
its
m
atch
in
g
class
ce
n
tr
e
co
n
s
titu
te
th
e
f
u
n
d
am
e
n
tal
p
r
in
ci
p
les o
f
ce
n
t
r
e
lo
s
s
.
T
h
e
f
o
llo
win
g
is
th
e
f
o
r
m
u
la
f
o
r
ce
n
tr
e
l
o
s
s
:
L
ctr
=
1
2
∑
‖
−
=
1
‖
2
2
(
7
)
W
h
er
e
ℝ
r
ep
r
esen
ts
th
e
d
ee
p
f
ea
tu
r
e
o
f
th
e
j
-
th
th
at
b
elo
n
g
s
to
th
e
y
-
th
class
ℝ
ac
t
as
th
e
ce
n
tr
e
o
f
th
e
d
ee
p
f
ea
tu
r
es
f
o
r
th
e
y
j
-
th
class
an
d
d
r
ep
r
esen
ts
th
e
f
ea
tu
r
e
d
im
en
s
io
n
.
T
h
e
ce
n
tr
es
ar
e
m
o
d
if
ied
d
ep
en
d
i
n
g
o
n
t
h
e
m
in
i
-
b
atch
es
r
ath
er
th
an
th
e
co
m
p
lete
tr
ain
in
g
s
et
s
in
ce
it
is
im
p
o
s
s
ib
le
to
co
m
p
u
te
u
s
in
g
th
e
en
tire
tr
ain
in
g
s
et.
Ad
d
itio
n
ally
,
a
s
ca
lar
is
u
tili
ze
d
to
lim
it
th
e
ce
n
tr
es’
lear
n
in
g
r
ate
in
o
r
d
er
to
less
en
th
e
s
ev
er
e
d
is
r
u
p
tio
n
s
g
en
er
ated
b
y
a
s
m
all
n
u
m
b
er
o
f
m
is
lab
elled
s
am
p
les.
T
h
e
m
ain
p
u
r
p
o
s
e
o
f
b
r
ai
n
tu
m
o
r
d
etec
tio
n
is
to
d
is
tin
g
u
is
h
b
etwe
en
s
am
p
les
th
at
ar
e
n
o
r
m
al
an
d
th
o
s
e
th
at
ar
e
an
o
m
alo
u
s
.
I
n
o
r
d
e
r
to
co
m
p
u
te
th
e
ce
n
tr
e
f
o
r
a
s
in
g
l
e
class
alo
n
e,
th
e
n
u
m
b
er
o
f
c
lass
es
f
o
r
th
e
ce
n
tr
e
lo
s
s
was
lo
wer
ed
to
o
n
e.
T
o
y
ield
th
e
o
n
e
-
class
ce
n
tr
e
lo
s
s
,
th
e
E
u
clid
ea
n
d
is
tan
ce
b
etwe
en
th
e
f
ea
tu
r
e
ce
n
tr
e
an
d
th
e
d
ee
p
f
ea
tu
r
es
o
f
all
s
am
p
le
ty
p
es is
co
m
p
u
ted
.
T
h
e
f
o
r
m
u
la
f
o
r
co
m
p
u
tatio
n
s
is
as f
o
llo
ws
:
L
octr
=
1
2
∑
‖
−
=
1
‖
2
2
(
8
)
W
h
er
e
r
e
p
r
esen
ts
th
e
ce
n
tr
al
f
ea
tu
r
e
o
f
all
t
h
e
n
o
r
m
al
s
am
p
les.
T
h
e
m
ix
ed
ce
n
t
r
e
lo
s
s
,
o
b
tain
ed
b
y
th
e
co
m
b
in
atio
n
o
f
th
e
o
n
e
-
class
ce
n
tr
e
lo
s
s
with
th
at
o
f
th
e
s
tan
d
ar
d
ce
n
t
r
e
lo
s
s
,
is
as f
o
llo
ws:
L
mixctr
=λ
1
L
ctr
+ λ
2
L
octr
=
λ1
2
∑
‖
−
=
1
‖
2
2
+
λ2
2
∑
‖
−
=
1
‖
2
2
(
9
)
4
.
4
.
Rec
o
ns
t
ruct
io
n lo
s
s
a
nd
cla
s
s
if
ica
t
io
n lo
s
s
T
h
e
p
r
o
p
o
s
ed
n
etwo
r
k
m
o
d
el
in
co
r
p
o
r
ates
b
o
th
th
e
r
ec
o
n
s
tr
u
ctio
n
an
d
class
if
ic
atio
n
lo
s
s
s
im
u
ltan
eo
u
s
ly
.
I
n
o
r
d
er
t
o
ex
tr
ac
t
m
u
ltip
le
r
ec
o
n
s
tr
u
ctio
n
laten
t
r
ep
r
esen
tatio
n
s
,
th
e
n
etwo
r
k
u
s
es
two
p
ar
allel
r
ec
o
n
s
tr
u
ctio
n
b
r
an
ch
es
as
s
h
o
wn
in
Fig
u
r
e
1
.
T
h
e
m
ea
n
s
q
u
ar
e
d
er
r
o
r
u
s
ed
f
o
r
t
h
e
r
ec
o
n
s
tr
u
ctio
n
o
f
th
ese
two
b
r
an
ch
es is
as f
o
llo
ws:
L
rec1
=
1
∑
‖
−
′
=
1
‖
2
2
(
1
0
)
L
rec2
=
1
∑
‖
−
′′
=
1
‖
2
2
(
1
1
)
W
h
er
e
A
j
d
en
o
tes
th
e
in
p
u
t
im
ag
e,
A
j
d
en
o
tes
th
e
f
ir
s
t
b
r
a
n
ch
,
th
e
r
ec
o
n
s
tr
u
cted
im
ag
e
,
an
d
A
j
d
ep
icts
th
e
r
ec
o
n
s
tr
u
cted
im
ag
e
f
r
o
m
th
e
s
ec
o
n
d
b
r
an
ch
.
L
AdaCos
,
an
an
g
u
lar
m
ar
g
in
lo
s
s
,
is
u
s
ed
in
t
h
e
class
if
icatio
n
lo
s
s
[
3
2
]
.
T
o
en
f
o
r
ce
an
an
g
u
lar
m
ar
g
in
b
etwe
en
d
if
f
er
e
n
t
ca
teg
o
r
ies
an
d
m
in
im
ize
th
e
co
s
in
e
a
n
g
le
to
th
e
lear
n
ed
m
ea
n
o
f
k
n
o
wn
ca
teg
o
r
ies,
th
e
lo
s
s
Ad
aCo
s
p
r
o
m
o
tes
lo
w
in
tr
a
-
class
v
ar
ian
ce
an
d
h
ig
h
i
n
ter
-
class
v
ar
ian
ce
.
I
n
t
h
is
lo
s
s
,
th
e
p
r
o
b
ab
ilit
y
o
f
s
am
p
le
A
j
th
at
b
elo
n
g
s
to
t
h
e
j
th
ca
teg
o
r
y
,
am
o
n
g
all
ca
teg
o
r
ies
K,
is
g
iv
en
as:
P
i,
j
=
̃
.
,
∑
̃
.
,
=
1
(
1
2
)
W
h
er
e
i
,
j
is
o
b
tain
ed
b
y
co
s
in
e
s
im
ilar
ity
co
s
i,
j
=
⟨
,
⟩
/
‖
,
‖
‖
‖
,
ℝ
is
a
lear
n
ed
ca
teg
o
r
y
ce
n
tr
e.
T
h
e
d
y
n
am
ically
ad
a
p
tiv
e
s
ca
lin
g
p
ar
am
ete
r
̃
(
)
at
tr
ain
in
g
s
tep
∈
0
is
d
ef
in
ed
as:
̃
(
)
=
{
√
2
(
−
1
)
(
)
(
(
4
,
)
)
(
1
3
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J
I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
4
5
9
-
1
4
6
8
1464
wh
er
e
(t)
med
[
0
,
]
r
ep
r
esen
ts
th
e
an
g
u
lar
ce
n
ter
o
f
all
th
e
co
r
r
esp
o
n
d
in
g
ca
teg
o
r
ies
th
at
ar
e
u
s
ed
in
th
is
cu
r
r
en
t
b
atch
.
T
h
e
d
ef
in
itio
n
o
f
(
)
is
ex
p
r
ess
ed
as:
(
)
=
1
∑
∑
ⅇ
−
−
1
,
c
os
,
≠
∈
(
)
(
1
4
)
W
h
er
e
N
(l)
r
ep
r
esen
ts
th
e
in
d
ic
es in
th
e
m
in
i
-
b
atch
at
th
e
t
-
th
iter
atio
n
.
ⅇ
−
−
1
,
c
os
,
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
lo
s
s
f
u
n
c
tio
n
co
n
s
is
ts
o
f
:
L
total
=
L
rec1
+ L
rec2
+
L
AdaCos
+
L
mixctr
(
1
5
)
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
ex
p
er
im
en
t
is
c
o
n
d
u
cted
o
n
th
e
d
ataset,
wh
ich
is
av
ailab
le
to
th
e
p
u
b
lic
f
r
o
m
a
n
o
p
en
-
s
o
u
r
ce
p
latf
o
r
m
ca
lled
“Ka
g
g
le”
[
3
7
]
.
T
h
e
d
ataset
co
n
tain
s
3
2
6
4
MRI
im
ag
es
s
p
li
t
in
to
two
g
r
o
u
p
s
o
f
2
8
7
0
an
d
8
7
0
im
ag
es
f
o
r
tr
ain
in
g
an
d
test
in
g
,
r
esp
ec
tiv
ely
,
an
d
a
r
e
in
tu
r
n
class
if
ied
in
to
f
o
u
r
cl
ass
es
:
m
en
in
g
io
m
a,
p
itu
itar
y
,
g
lio
m
a
,
an
d
n
o
t
u
m
o
r
.
Fig
u
r
e
4
d
e
p
icts
s
am
p
le
im
ag
es
o
f
th
e
f
o
u
r
class
es.
Pr
ep
r
o
ce
s
s
in
g
o
f
d
ata
im
ag
es
is
d
o
n
e
b
y
r
esizin
g
th
e
tr
ain
in
g
an
d
test
in
g
im
ag
es
to
a
f
ix
ed
s
ize,
s
h
u
f
f
lin
g
th
ese
im
ag
e
s
,
en
co
d
in
g
th
e
lab
els
wh
er
e
ea
ch
lab
el
i
s
m
ap
p
ed
to
a
u
n
iq
u
e
in
te
g
e
r
v
alu
e,
an
d
d
o
in
g
d
ata
au
g
m
en
tatio
n
.
Py
th
o
n
’
s
T
en
s
o
r
Flo
w
an
d
Ker
as
ar
e
u
s
ed
to
p
er
f
o
r
m
th
e
ex
p
er
im
e
n
t
o
n
a
C
o
r
e
i5
C
P
U
with
8
G
B
R
AM
r
u
n
n
in
g
at
2
.
4
2
GHz
.
T
h
e
m
o
d
el
was tr
ain
ed
f
o
r
2
0
0
ep
o
ch
s
.
T
ab
le
1
p
r
esen
ts
th
e
b
r
ain
t
u
m
o
r
class
if
icatio
n
r
esu
lts
o
f
1
0
DL
m
o
d
els.
Am
o
n
g
th
e
m
,
t
h
e
SS
DH
R
class
if
ier
n
etwo
r
k
ac
h
iev
ed
th
e
b
est
p
er
f
o
r
m
an
ce
with
9
0
.
2
%
ac
cu
r
ac
y
,
alo
n
g
with
th
e
h
ig
h
est
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
an
d
Kap
p
a
s
co
r
e.
Acc
u
r
ac
y
r
ep
r
esen
ts
th
e
o
v
er
all
co
r
r
ec
t
p
r
e
d
ictio
n
s
,
p
r
ec
is
io
n
in
d
icate
s
p
r
ed
ictio
n
r
e
liab
ilit
y
,
r
ec
all
m
ea
s
u
r
es
th
e
ab
ilit
y
to
id
en
tify
p
o
s
itiv
e
ca
s
es,
F1
-
s
co
r
e
b
alan
ce
s
p
r
ec
is
io
n
an
d
r
ec
all,
an
d
th
e
Kap
p
a
s
co
r
e
ev
alu
ates
th
e
a
g
r
ee
m
en
t
b
et
wee
n
p
r
e
d
icted
a
n
d
ac
t
u
al
v
alu
es.
T
h
ese
r
esu
lts
d
em
o
n
s
tr
ate
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
SS
DHR
class
if
ier
n
etwo
r
k
in
lear
n
in
g
c
o
m
p
le
x
p
atte
r
n
s
f
o
r
b
r
ain
tu
m
o
r
class
if
icatio
n
.
T
ab
le
2
s
h
o
ws
th
e
m
ac
r
o
-
av
er
ag
e
d
d
ia
g
n
o
s
t
ic
p
er
f
o
r
m
an
ce
o
f
d
if
f
er
en
t
C
NN
m
o
d
els
u
s
in
g
p
r
ec
is
io
n
,
r
ec
all,
F1
s
co
r
e
,
an
d
AUC
m
etr
ics.
Ma
cr
o
-
av
er
a
g
e
is
ca
lcu
lated
b
y
a
v
er
ag
in
g
th
e
m
etr
ic
v
al
u
es
ac
r
o
s
s
all
class
es
eq
u
ally
.
Hi
g
h
er
m
ac
r
o
-
av
er
ag
e
v
al
u
es
in
d
icate
b
etter
o
v
er
all
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
Ma
c
r
o
-
av
er
ag
e
p
r
ec
is
io
n
,
r
ec
all,
F1
s
co
r
e,
an
d
AUC
ar
e
o
b
tain
ed
b
y
av
er
a
g
in
g
t
h
e
co
r
r
esp
o
n
d
in
g
v
alu
es
ca
lcu
lated
f
o
r
ea
ch
class
.
Fig
u
r
e
4
.
Sam
p
le
im
a
g
e
o
f
ea
c
h
class
f
r
o
m
th
e
d
ataset
T
ab
le
1
.
Ov
e
r
all
m
etr
ics o
f
all
m
o
d
els
M
o
d
e
l
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F
1
sc
o
r
e
K
a
p
p
a
I
n
c
e
p
t
i
o
n
-
r
e
s
n
e
t
v
2
[
2
8
]
2
9
.
3
8
.
6
2
9
.
3
1
.
0
0
Le
n
e
t
5
[
2
9
]
74
29
2
7
.
1
2
6
.
9
5
.
7
R
e
sN
e
t
1
0
1
-
LST
M
[
3
0
]
7
5
.
4
8
0
.
1
7
5
.
4
7
3
.
7
6
5
.
8
I
n
c
e
p
t
i
o
n
v
3
[
3
1
]
7
6
.
4
7
7
.
5
7
6
.
4
94
68
X
c
e
p
t
i
o
n
[
3
2
]
7
6
.
6
7
6
.
4
7
6
.
7
9
2
.
3
6
8
.
5
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
[
3
3
]
7
7
.
4
8
4
.
8
7
6
.
7
7
4
.
6
6
1
.
3
C
o
n
v
N
e
X
t
B
a
s
e
[
3
4
]
86
8
7
.
8
8
7
.
1
9
8
.
5
8
2
.
5
V
g
g
1
9
[
3
5
]
8
6
.
8
86
86
9
5
.
4
8
2
.
2
R
e
sN
e
t
1
0
1
[
3
6
]
8
8
.
6
89
8
8
.
9
9
6
.
1
8
4
.
6
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
9
1
.
9
9
2
.
6
9
2
.
3
9
7
.
5
8
9
.
6
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
S
S
DHR
cla
s
s
ifier
n
etw
o
r
k
fo
r
b
r
a
in
tu
mo
r
d
etec
tio
n
u
s
in
g
MRI
ima
g
es
(
K
a
lp
a
n
a
Telka
r
)
1465
T
ab
le
2
.
Ma
cr
o
av
er
a
g
e
M
o
d
e
l
P
r
e
c
i
s
i
o
n
r
e
c
a
l
l
F
1
sc
o
r
e
AUC
I
n
c
e
p
t
i
o
n
-
r
e
s
n
e
t
v
2
[
2
8
]
7
25
11
50
Le
n
e
t
5
[
2
9
]
24
24
22
52
R
e
sN
e
t
1
0
1
-
LST
M
[
3
0
]
80
75
74
8
2
.
2
I
n
c
e
p
t
i
o
n
v
3
[
3
1
]
78
76
76
84
X
c
e
p
t
i
o
n
[
3
2
]
77
77
77
8
4
.
6
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
[
3
3
]
85
77
75
9
2
.
4
C
o
n
v
N
e
X
t
B
a
s
e
[
3
4
]
89
88
88
9
1
.
7
V
g
g
1
9
[
3
5
]
87
87
87
9
1
.
4
R
e
sN
e
t
1
0
1
[
3
6
]
89
89
89
92
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
92
93
92
9
5
.
1
T
ab
le
3
s
h
o
ws
th
e
m
icr
o
-
av
er
ag
e
o
f
v
ar
io
u
s
DL
m
o
d
els.
T
h
is
m
o
d
el,
wh
ich
h
as
th
e
h
ig
h
est
m
icr
o
-
av
er
ag
e
i
n
all
th
e
s
et
m
etr
ics,
is
co
n
s
id
er
ed
t
h
e
b
est
-
p
e
r
f
o
r
m
in
g
m
o
d
el.
Mic
r
o
-
av
er
a
g
e
p
r
e
cisi
o
n
is
co
m
p
u
te
d
b
y
co
n
s
id
er
in
g
all
tr
u
e
p
o
s
itiv
e
(
T
P)
an
d
f
alse p
o
s
itiv
e
(
FP
)
p
r
ed
ictio
n
s
ac
r
o
s
s
all
class
es
t
o
g
et
h
er
,
r
ath
er
th
an
co
m
p
u
tin
g
p
r
ec
is
io
n
p
er
class
in
d
iv
id
u
ally
.
Mic
r
o
-
av
er
a
g
e
r
ec
all
is
co
m
p
u
ted
b
y
s
u
m
m
in
g
th
e
tr
u
e
p
o
s
itiv
es
(
T
P)
f
o
r
all
class
es
an
d
d
iv
id
i
n
g
b
y
th
e
s
u
m
o
f
th
e
tr
u
e
p
o
s
i
tiv
es
(
T
P)
an
d
f
alse
n
eg
ativ
es
(
FN)
f
o
r
all
class
es.
T
h
e
m
icr
o
-
av
e
r
ag
e
F1
-
s
co
r
e
i
s
th
e
h
ar
m
o
n
ic
m
ea
n
o
f
th
e
m
icr
o
-
av
er
a
g
e
p
r
ec
is
io
n
an
d
m
i
cr
o
-
av
er
a
g
e
r
ec
all.
T
h
e
m
icr
o
-
av
er
a
g
e
AUC
is
d
ef
in
ed
as
th
e
AUC
o
f
th
e
R
OC
g
r
ap
h
co
m
p
u
ted
f
r
o
m
th
e
tr
u
e
p
o
s
itiv
e
an
d
f
alse
p
o
s
itiv
e
r
atio
s
o
f
all
class
es.
T
ab
le
4
p
r
esen
ts
weig
h
ted
av
er
ag
e
m
etr
ics
u
s
ed
to
ev
alu
ate
m
o
d
el
p
er
f
o
r
m
an
ce
,
esp
e
cially
f
o
r
im
b
alan
ce
d
d
atasets
.
W
eig
h
ted
av
er
a
g
e
p
r
ec
is
io
n
is
ca
lcu
lated
b
y
av
er
ag
i
n
g
th
e
p
r
ec
is
io
n
o
f
ea
ch
class
b
ase
d
o
n
th
e
n
u
m
b
er
o
f
tr
u
e
in
s
tan
ce
s
in
th
at
class
.
Similar
ly
,
wei
g
h
ted
av
er
ag
e
r
ec
all
is
th
e
av
er
ag
e
r
ec
all
weig
h
ted
b
y
th
e
tr
u
e
in
s
tan
ce
s
o
f
ea
ch
class
.
T
h
e
weig
h
ted
av
er
ag
e
F
1
s
co
r
e
is
o
b
tain
ed
b
y
a
v
er
ag
i
n
g
th
e
F1
s
co
r
es
o
f
all
class
e
s
u
s
in
g
th
e
n
u
m
b
er
o
f
tr
u
e
s
am
p
les
as
weig
h
ts
.
W
eig
h
ted
av
er
ag
e
AUC,
h
o
wev
er
,
is
g
en
e
r
ally
n
o
t
in
clu
d
ed
in
s
tan
d
ar
d
class
if
icatio
n
r
ep
o
r
ts
.
T
ab
le
5
d
e
p
icts
th
e
co
n
f
u
s
io
n
m
atr
ix
f
o
r
ea
ch
m
o
d
el
in
t
h
e
class
if
icatio
n
.
I
n
th
e
co
n
f
u
s
io
n
m
atr
ix
,
ea
ch
o
f
th
e
r
o
ws
c
o
r
r
esp
o
n
d
s
to
th
e
t
r
u
e
class
es,
wh
ile
ea
c
h
o
f
th
e
co
lu
m
n
s
co
r
r
esp
o
n
d
s
to
t
h
e
p
r
ed
ic
ted
class
es.
T
h
e
n
u
m
b
er
n
ex
t
to
ea
ch
ce
ll
r
ep
r
esen
ts
th
e
n
u
m
b
er
o
f
s
am
p
les
th
at
wer
e
class
if
ied
co
r
r
ec
tl
y
,
in
co
r
r
ec
tly
,
o
r
m
is
class
if
ied
.
T
h
is
tab
le
s
u
m
m
ar
izes
th
e
f
in
d
i
n
g
s
o
n
th
e
p
er
f
o
r
m
an
ce
o
f
v
ar
io
u
s
m
o
d
els
in
ac
cu
r
ately
p
er
f
o
r
m
in
g
th
e
class
if
icatio
n
ta
s
k
o
f
tu
m
o
r
ty
p
es.
T
a
b
le
6
d
is
p
lay
s
th
e
s
en
s
itiv
ity
o
f
ea
ch
m
o
d
el
f
o
r
ea
ch
class
.
R
ec
all,
also
k
n
o
wn
as
s
en
s
itiv
ity
,
is
d
ef
in
ed
as
th
e
m
o
d
el'
s
ab
ilit
y
to
co
r
r
ec
tly
id
e
n
tify
th
e
p
o
s
itiv
e
s
am
p
les
(
tu
m
o
r
ty
p
es).
Hig
h
er
s
en
s
itiv
ity
v
alu
es
in
d
icate
th
at
th
e
m
o
d
el
is
b
etter
at
c
o
r
r
ec
tly
id
en
tify
in
g
tu
m
o
r
s
o
f
a
s
p
ec
if
ic
ty
p
e.
T
ab
le
3
.
Mic
r
o
av
er
a
g
e
M
o
d
e
l
P
r
e
c
i
s
i
o
n
r
e
c
a
l
l
F
1
sco
r
e
AUC
I
n
c
e
p
t
i
o
n
-
r
e
s
n
e
t
v
2
[
2
8
]
29
29
29
50
Le
n
e
t
5
[
2
9
]
2
9
.
5
2
9
.
5
2
9
.
5
51
R
e
sN
e
t
1
0
1
-
LST
M
[
3
0
]
58
58
58
8
2
.
7
I
n
c
e
p
t
i
o
n
v
3
[
3
1
]
7
6
.
4
7
6
.
4
7
6
.
4
93
X
c
e
p
t
i
o
n
[
3
2
]
7
6
.
7
7
6
.
7
7
6
.
7
9
3
.
1
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
[
3
3
]
7
7
.
4
7
7
.
4
7
7
.
4
8
9
.
9
C
o
n
v
N
e
X
t
B
a
s
e
[
3
4
]
8
7
.
1
8
7
.
1
8
7
.
1
9
7
.
1
V
g
g
1
9
[
3
5
]
8
6
.
8
8
6
.
8
8
6
.
8
9
3
.
9
R
e
sN
e
t
1
0
1
[
3
6
]
8
8
.
6
8
8
.
6
8
8
.
6
9
7
.
3
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
9
2
.
3
9
2
.
3
9
2
.
3
9
8
.
6
T
ab
le
4
.
W
eig
h
ted
a
v
er
ag
e
M
o
d
e
l
P
r
e
c
i
s
i
o
n
r
e
c
a
l
l
F
1
sco
r
e
AUC
I
n
c
e
p
t
i
o
n
-
r
e
s
n
e
t
v
2
[
2
8
]
9
29
13
50
Le
n
e
t
5
[
2
9
]
24
25
23
52
R
e
sN
e
t
1
0
1
-
LST
M
[
3
0
]
80
75
73
8
1
.
3
I
n
c
e
p
t
i
o
n
v
3
[
3
1
]
78
76
76
8
3
.
9
X
c
e
p
t
i
o
n
[
3
2
]
76
77
76
84
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
[
3
3
]
83
77
74
9
2
.
1
C
o
n
v
N
e
X
t
B
a
s
e
[
3
4
]
88
87
87
9
1
.
1
V
g
g
1
9
[
3
5
]
87
87
87
9
1
.
1
R
e
sN
e
t
1
0
1
[
3
6
]
89
89
89
9
2
.
3
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
93
92
92
9
4
.
9
T
ab
le
5
.
C
o
n
f
u
s
io
n
m
atr
ix
f
o
r
all
m
o
d
els
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
C
o
n
v
N
e
X
t
B
a
s
e
V
g
g
1
9
R
e
sN
e
t
1
0
1
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J
I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
4
5
9
-
1
4
6
8
1466
T
ab
le
6
.
Sen
s
itiv
ity
/R
ec
all
f
o
r
ea
ch
in
d
iv
i
d
u
al
class
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
C
o
n
v
N
e
X
t
B
a
s
e
V
g
g
1
9
R
e
sN
e
t
1
0
1
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
T
ab
le
7
illu
s
tr
ates
th
e
s
elec
tiv
ity
o
f
ea
ch
m
o
d
el
f
o
r
ea
ch
cla
s
s
.
Pre
cisi
o
n
,
also
k
n
o
wn
as
s
elec
tiv
ity
,
is
d
ef
in
ed
as
a
m
o
d
el'
s
ab
ilit
y
to
co
r
r
ec
tly
ex
clu
d
e
n
eg
ati
v
e
s
am
p
les
(
s
am
p
les
th
at
ar
e
n
o
t
o
f
th
e
s
p
ec
if
ie
d
tu
m
o
r
ty
p
e
)
.
Hig
h
er
s
elec
tiv
it
y
v
alu
es su
g
g
est th
a
t th
e
m
o
d
e
l is b
etter
at
co
r
r
ec
tly
id
en
tify
i
n
g
s
am
p
les th
at
ar
e
n
o
t
o
f
th
e
s
p
ec
if
ic
tu
m
o
r
ty
p
e.
T
ab
le
8
p
r
esen
ts
th
e
p
r
ec
i
s
io
n
–
r
ec
all
cu
r
v
es
o
f
all
m
o
d
els,
illu
s
tr
atin
g
th
e
tr
ad
e
-
o
f
f
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all
at
d
if
f
er
en
t th
r
esh
o
ld
v
alu
es.
Mo
d
el
p
er
f
o
r
m
an
ce
is
ev
alu
ated
u
s
in
g
th
e
ar
ea
u
n
d
e
r
th
e
c
u
r
v
e
(
AUC),
wh
er
e
a
h
ig
h
er
AUC in
d
icate
s
b
etter
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
Am
o
n
g
all
m
o
d
els,
th
e
SS
DHR
cla
s
s
if
ier
n
etwo
r
k
ac
h
iev
ed
th
e
b
est
r
esu
lts
wi
th
9
1
.
9
%
ac
cu
r
ac
y
,
9
2
.
6
%
p
r
ec
is
io
n
,
9
2
.
3
%
r
ec
al
l,
9
7
.
5
%
F1
-
sc
o
r
e,
8
9
.
6
%
Ka
p
p
a
s
co
r
e,
9
5
.
1
m
ac
r
o
-
av
e
r
ag
e
AUC,
an
d
9
8
.
6
m
icr
o
-
av
er
a
g
e
AUC.
I
t
also
o
b
tain
ed
t
h
e
h
i
g
h
est
weig
h
te
d
av
er
a
g
e
m
etr
ics
alo
n
g
with
s
u
p
er
io
r
s
en
s
itiv
ity
an
d
s
p
ec
if
icity
f
o
r
ea
ch
clas
s
.
T
h
ese
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
SS
DHR
class
if
ier
n
etwo
r
k
is
th
e
m
o
s
t
ef
f
ec
tiv
e
m
o
d
el
f
o
r
b
r
ain
t
u
m
o
r
class
if
icatio
n
co
m
p
ar
ed
t
o
ex
is
tin
g
ap
p
r
o
ac
h
es.
T
ab
le
7
.
Selectiv
ity
/p
r
ec
is
io
n
s
f
o
r
ea
c
h
in
d
iv
i
d
u
al
class
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
C
o
n
v
N
e
X
t
B
a
s
e
V
g
g
1
9
R
e
sN
e
t
1
0
1
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
T
ab
le
8
.
Pre
cisi
o
n
-
r
ec
all
c
u
r
v
e
R
e
sN
e
t
1
0
1
-
LST
M
-
G
R
U
C
o
n
v
N
e
X
t
B
a
s
e
V
g
g
1
9
R
e
sN
e
t
1
0
1
S
S
D
H
R
c
l
a
ssi
f
i
e
r
n
e
t
w
o
r
k
6.
CO
NCLU
SI
O
N
T
h
e
p
r
o
p
o
s
ed
SS
DHR
clas
s
if
ier
n
etwo
r
k
m
o
d
el
u
s
ed
f
o
r
p
er
f
o
r
m
in
g
th
e
b
r
ain
tu
m
o
r
class
if
icatio
n
task
is
co
m
p
ar
ed
with
o
t
h
er
s
tate
-
of
-
th
e
-
ar
t
C
NN
m
o
d
els.
E
x
p
er
im
e
n
tal
r
esu
lts
illu
s
tr
at
e
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
o
u
tp
er
f
o
r
m
s
e
x
is
tin
g
m
o
d
els
in
v
ar
io
u
s
m
etr
ics,
ac
h
iev
in
g
an
a
cc
u
r
ac
y
o
f
9
1
.
9
%
o
n
th
e
b
r
ain
tu
m
o
r
class
if
icatio
n
d
ataset.
B
ased
o
n
th
ese
f
in
d
in
g
s
,
it
is
co
n
clu
d
ed
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
is
m
o
r
e
ef
f
ec
tiv
e
in
d
iag
n
o
s
in
g
f
o
u
r
d
is
tin
ct
ty
p
es
o
f
b
r
ain
tu
m
o
r
s
at
an
ea
r
ly
s
tag
e.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
lear
n
s
in
tr
icate
f
ea
tu
r
es
f
r
o
m
th
e
im
ag
es o
f
t
h
e
MRI
s
ca
n
s
,
wh
ich
aid
s
in
im
p
r
o
v
in
g
t
h
e
ef
f
icien
cy
an
d
ac
c
u
r
ac
y
o
f
th
e
m
o
d
el.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
th
o
r
s
s
tate
th
at
n
o
f
u
n
d
i
n
g
was
in
v
o
lv
e
d
.
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
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
S
S
DHR
cla
s
s
ifier
n
etw
o
r
k
fo
r
b
r
a
in
tu
mo
r
d
etec
tio
n
u
s
in
g
MRI
ima
g
es
(
K
a
lp
a
n
a
Telka
r
)
1467
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Kalp
an
a
T
elk
ar
✓
✓
✓
✓
✓
✓
K.
An
u
s
u
d
h
a
✓
✓
✓
✓
✓
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
s
tate
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ataset
u
s
ed
in
th
is
m
an
u
s
cr
ip
t
is
av
ailab
le
p
u
b
licly
,
an
d
a
r
ef
er
en
ce
was
g
iv
en
in
th
e
R
esu
lt
s
an
d
Dis
cu
s
s
io
n
s
ec
tio
n
.
RE
F
E
R
E
NC
E
S
[
1
]
H
.
H
.
S
u
l
t
a
n
,
N
.
M
.
S
a
l
e
m
,
a
n
d
W
.
A
l
-
A
t
a
b
a
n
y
,
“
M
u
l
t
i
-
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
b
r
a
i
n
t
u
m
o
r
i
m
a
g
e
s
u
s
i
n
g
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
EE
E
Ac
c
e
ss
,
v
o
l
.
7
,
p
p
.
6
9
2
1
5
–
6
9
2
2
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
a
c
c
e
ss
.
2
0
1
9
.
2
9
1
9
1
2
2
.
[
2
]
M
.
I
.
R
a
z
z
a
k
,
M
.
I
mr
a
n
,
a
n
d
G
.
X
u
,
“
Ef
f
i
c
i
e
n
t
b
r
a
i
n
t
u
m
o
r
se
g
m
e
n
t
a
t
i
o
n
w
i
t
h
m
u
l
t
i
sc
a
l
e
t
w
o
-
p
a
t
h
w
a
y
-
g
r
o
u
p
c
o
n
v
e
n
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
I
EEE
J
o
u
rn
a
l
o
f
Bi
o
m
e
d
i
c
a
l
a
n
d
H
e
a
l
t
h
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
2
3
,
n
o
.
5
,
p
p
.
1
9
1
1
–
1
9
1
9
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
j
b
h
i
.
2
0
1
8
.
2
8
7
4
0
3
3
.
[
3
]
R
.
L.
S
i
e
g
e
l
,
K
.
D
.
M
i
l
l
e
r
,
a
n
d
A
.
J
e
mal
,
“
C
a
n
c
e
r
st
a
t
i
s
t
i
c
s,
2
0
1
7
,
”
C
A:
A
C
a
n
c
e
r
J
o
u
r
n
a
l
f
o
r
C
l
i
n
i
c
i
a
n
s
,
v
o
l
.
6
7
,
n
o
.
1
,
p
p
.
7
–
3
0
,
Jan
.
2
0
1
7
,
d
o
i
:
1
0
.
3
3
2
2
/
c
a
a
c
.
2
1
3
8
7
.
[
4
]
Y
.
Zh
a
n
g
,
A
.
L
i
,
C
.
P
e
n
g
,
a
n
d
M
.
W
a
n
g
,
“
I
mp
r
o
v
e
g
l
i
o
b
l
a
st
o
m
a
m
u
l
t
i
f
o
r
me
p
r
o
g
n
o
si
s
p
r
e
d
i
c
t
i
o
n
b
y
u
si
n
g
f
e
a
t
u
r
e
se
l
e
c
t
i
o
n
a
n
d
mu
l
t
i
p
l
e
k
e
r
n
e
l
l
e
a
r
n
i
n
g
,
”
I
EEE
/
AC
M
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
a
t
i
o
n
a
l
B
i
o
l
o
g
y
a
n
d
Bi
o
i
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
1
3
,
n
o
.
5
,
p
p
.
8
2
5
–
8
3
5
,
2
0
1
6
,
d
o
i
:
1
0
.
1
1
0
9
/
t
c
b
b
.
2
0
1
6
.
2
5
5
1
7
4
5
.
[
5
]
W
.
A
y
a
d
i
,
W
.
E
l
h
a
mz
i
,
I
.
C
h
a
r
f
i
,
a
n
d
M
.
A
t
r
i
,
“
A
h
y
b
r
i
d
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
a
p
p
r
o
a
c
h
f
o
r
b
r
a
i
n
M
R
I
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
a
se
d
o
n
b
a
g
-
of
-
w
o
r
d
s
,
”
Bi
o
m
e
d
i
c
a
l
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
4
8
,
p
p
.
1
4
4
–
1
5
2
,
F
e
b
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
sp
c
.
2
0
1
8
.
1
0
.
0
1
0
.
[
6
]
T.
K
a
u
r
,
B
.
S
.
S
a
i
n
i
,
a
n
d
S
.
G
u
p
t
a
,
“
Q
u
a
n
t
i
t
a
t
i
v
e
met
r
i
c
f
o
r
M
R
b
r
a
i
n
t
u
mo
u
r
g
r
a
d
e
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
s
a
m
p
l
e
s
p
a
c
e
d
e
n
s
i
t
y
mea
s
u
r
e
o
f
a
n
a
l
y
t
i
c
i
n
t
r
i
n
s
i
c
mo
d
e
f
u
n
c
t
i
o
n
r
e
p
r
e
se
n
t
a
t
i
o
n
,
”
I
ET
I
m
a
g
e
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
1
,
n
o
.
8
,
p
p
.
6
2
0
–
6
3
2
,
2
0
1
7
,
d
o
i
:
1
0
.
1
0
4
9
/
i
e
t
-
i
p
r
.
2
0
1
6
.
1
1
0
3
.
[
7
]
K
.
U
sm
a
n
a
n
d
K
.
R
a
j
p
o
o
t
,
“
B
r
a
i
n
t
u
mo
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
f
r
o
m
m
u
l
t
i
-
m
o
d
a
l
i
t
y
M
R
I
u
si
n
g
w
a
v
e
l
e
t
s
a
n
d
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
P
a
t
t
e
r
n
An
a
l
y
si
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
0
,
n
o
.
3
,
p
p
.
8
7
1
–
8
8
1
,
F
e
b
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
0
4
4
-
017
-
0
5
9
7
-
8.
[
8
]
V
.
A
n
i
t
h
a
a
n
d
S
.
M
u
r
u
g
a
v
a
l
l
i
,
“
B
r
a
i
n
t
u
mo
u
r
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
t
w
o
‐
t
i
e
r
c
l
a
ss
i
f
i
e
r
w
i
t
h
a
d
a
p
t
i
v
e
s
e
g
m
e
n
t
a
t
i
o
n
t
e
c
h
n
i
q
u
e
,
”
I
ET
C
o
m
p
u
t
e
r
Vi
s
i
o
n
,
v
o
l
.
1
0
,
n
o
.
1
,
p
p
.
9
–
1
7
,
F
e
b
.
2
0
1
6
,
d
o
i
:
1
0
.
1
0
4
9
/
i
e
t
-
c
v
i
.
2
0
1
4
.
0
1
9
3
.
[
9
]
E.
-
S
.
A
.
E
l
-
D
a
h
s
h
a
n
,
H
.
M
.
M
o
h
se
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
m
a
n
b
r
a
i
n
t
u
m
o
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
.
[
1
0
]
A
.
P
a
s
h
a
e
i
,
H
.
S
a
j
e
d
i
,
a
n
d
N
.
Ja
z
a
y
e
r
i
,
“
B
r
a
i
n
t
u
m
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
v
i
a
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
a
n
d
e
x
t
r
e
me
l
e
a
r
n
i
n
g
mac
h
i
n
e
s
,
”
i
n
2
0
1
8
8
t
h
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
C
o
m
p
u
t
e
r
a
n
d
K
n
o
w
l
e
d
g
e
E
n
g
i
n
e
e
ri
n
g
(
I
C
C
K
E)
,
O
c
t
.
2
0
1
8
,
p
p
.
3
1
4
–
3
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
i
c
c
k
e
.
2
0
1
8
.
8
5
6
6
5
7
1
.
[
1
1
]
S
.
D
a
s,
O
.
F
.
M
.
R
.
R
.
A
r
a
n
y
a
,
a
n
d
N
.
N
.
La
b
i
b
a
,
“
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
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
,
”
M
a
y
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
ER
T.
2
0
1
9
.
8
9
3
4
6
0
3
.
[
1
2
]
S
.
D
e
e
p
a
k
a
n
d
P
.
M
.
A
m
e
e
r
,
“
B
r
a
i
n
t
u
m
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
d
e
e
p
C
N
N
f
e
a
t
u
r
e
s
v
i
a
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
,
”
C
o
m
p
u
t
e
rs
i
n
B
i
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
1
1
1
,
p
.
1
0
3
3
4
5
,
A
u
g
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
b
i
o
m
e
d
.
2
0
1
9
.
1
0
3
3
4
5
.
[
1
3
]
M
.
M
.
B
a
d
ž
a
a
n
d
M
.
Č
.
B
a
r
j
a
k
t
a
r
o
v
i
ć
,
“
C
l
a
s
si
f
i
c
a
t
i
o
n
o
f
b
r
a
i
n
t
u
mo
r
s
f
r
o
m
M
R
I
i
ma
g
e
s
u
si
n
g
a
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
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
0
,
n
o
.
6
,
p
.
1
9
9
9
,
M
a
r
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
0
0
6
1
9
9
9
.
[
1
4
]
I
.
A
b
d
El
K
a
d
e
r
,
G
.
X
u
,
Z
.
S
h
u
a
i
,
S
.
S
a
mi
n
u
,
I
.
Jav
a
i
d
,
a
n
d
I
.
S
a
l
i
m
A
h
mad
,
“
D
i
f
f
e
r
e
n
t
i
a
l
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
mo
d
e
l
f
o
r
b
r
a
i
n
t
u
m
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
Bra
i
n
S
c
i
e
n
c
e
s
,
v
o
l
.
1
1
,
n
o
.
3
,
p
.
3
5
2
,
M
a
r
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
b
r
a
i
n
s
c
i
1
1
0
3
0
3
5
2
.
[
1
5
]
N
.
B
h
a
r
d
w
a
j
,
M
.
S
o
o
d
,
a
n
d
S
.
S
.
G
i
l
l
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
r
a
m
e
w
o
r
k
u
si
n
g
C
N
N
f
o
r
b
r
a
i
n
t
u
mo
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
2
0
2
2
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Mu
l
t
i
m
e
d
i
a
,
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
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
i
e
s
(
I
M
PAC
T
)
,
N
o
v
.
2
0
2
2
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
i
m
p
a
c
t
5
5
5
1
0
.
2
0
2
2
.
1
0
0
2
9
0
4
3
.
[
1
6
]
M
.
S
a
j
j
a
d
,
S
.
K
h
a
n
,
K
.
M
u
h
a
m
ma
d
,
W
.
W
u
,
A
.
U
l
l
a
h
,
a
n
d
S
.
W
.
B
a
i
k
,
“
M
u
l
t
i
-
g
r
a
d
e
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
d
e
e
p
C
N
N
w
i
t
h
e
x
t
e
n
s
i
v
e
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
,
”
J
o
u
r
n
a
l
o
f
C
o
m
p
u
t
a
t
i
o
n
a
l
S
c
i
e
n
c
e
,
v
o
l
.
3
0
,
p
p
.
1
7
4
–
1
8
2
,
Jan
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
o
c
s
.
2
0
1
8
.
1
2
.
0
0
3
.
[
1
7
]
R
.
V
a
n
k
d
o
t
h
u
,
M
.
A
.
H
a
m
e
e
d
,
a
n
d
H
.
F
a
t
i
ma,
“
A
b
r
a
i
n
t
u
m
o
r
i
d
e
n
t
i
f
i
c
a
t
i
o
n
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
b
a
s
e
d
o
n
C
N
N
-
LSTM
me
t
h
o
d
,
”
C
o
m
p
u
t
e
rs
a
n
d
El
e
c
t
ri
c
a
l
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
0
1
,
p
.
1
0
7
9
6
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
e
l
e
c
e
n
g
.
2
0
2
2
.
1
0
7
9
6
0
.
[
1
8
]
W
.
A
y
a
d
i
,
W
.
El
h
a
mz
i
,
I
.
C
h
a
r
f
i
,
a
n
d
M
.
A
t
r
i
,
“
D
e
e
p
C
N
N
f
o
r
b
r
a
i
n
t
u
mo
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
N
e
u
r
a
l
Pr
o
c
e
ssi
n
g
L
e
t
t
e
rs
,
v
o
l
.
5
3
,
n
o
.
1
,
p
p
.
6
7
1
–
7
0
0
,
Ja
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
6
3
-
0
2
0
-
1
0
3
9
8
-
2.
[
1
9
]
E.
I
r
mak
,
“
M
u
l
t
i
-
c
l
a
s
si
f
i
c
a
t
i
o
n
o
f
b
r
a
i
n
t
u
m
o
r
M
R
I
i
ma
g
e
s
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
w
i
t
h
f
u
l
l
y
o
p
t
i
mi
z
e
d
f
r
a
mew
o
r
k
,
”
I
r
a
n
i
a
n
J
o
u
r
n
a
l
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
-
T
ra
n
s
a
c
t
i
o
n
s
o
f
El
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
4
5
,
n
o
.
3
,
p
p
.
1
0
1
5
–
1
0
3
6
,
A
p
r
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
0
9
9
8
-
0
2
1
-
0
0
4
2
6
-
9.
[
2
0
]
I
.
S
.
R
a
j
p
u
t
,
A
.
G
u
p
t
a
,
V
.
Ja
i
n
,
a
n
d
S
.
Ty
a
g
i
,
“
A
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
-
b
a
se
d
b
r
a
i
n
t
u
mo
r
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
ma
g
n
e
t
i
c
r
e
so
n
a
n
c
e
i
ma
g
e
s,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
3
,
n
o
.
7
,
p
p
.
2
0
4
8
7
–
2
0
5
0
6
,
A
u
g
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
0
2
3
-
1
6
1
4
3
-
w.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J
I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
4
5
9
-
1
4
6
8
1468
[
2
1
]
N
.
A
b
i
w
i
n
a
n
d
a
,
M
.
H
a
n
i
f
,
S
.
T.
H
e
s
a
p
u
t
r
a
,
A
.
H
a
n
d
a
y
a
n
i
,
a
n
d
T.
R
.
M
e
n
g
k
o
,
“
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
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
n
W
o
rl
d
C
o
n
g
ress
o
n
Me
d
i
c
a
l
Ph
y
si
c
s a
n
d
Bi
o
m
e
d
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
2
0
1
8
,
S
p
r
i
n
g
e
r
N
a
t
u
r
e
S
i
n
g
a
p
o
r
e
,
2
0
1
8
,
p
p
.
1
8
3
–
1
8
9
.
[
2
2
]
J.
S
.
P
a
u
l
,
A
.
J.
P
l
a
ss
a
r
d
,
B
.
A
.
La
n
d
man
,
a
n
d
D
.
F
a
b
b
r
i
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
b
r
a
i
n
t
u
m
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
M
e
d
i
c
a
l
I
m
a
g
i
n
g
2
0
1
7
:
Bi
o
m
e
d
i
c
a
l
A
p
p
l
i
c
a
t
i
o
n
s
i
n
Mo
l
e
c
u
l
a
r,
S
t
ru
c
t
u
r
a
l
,
a
n
d
F
u
n
c
t
i
o
n
a
l
I
m
a
g
i
n
g
,
M
a
r
.
2
0
1
7
,
v
o
l
.
1
0
1
3
7
,
p
.
1
0
1
3
7
1
0
,
d
o
i
:
1
0
.
1
1
1
7
/
1
2
.
2
2
5
4
1
9
5
.
[
2
3
]
W
.
Y
a
n
g
e
t
a
l
.
,
“
C
o
n
t
e
n
t
‐
b
a
se
d
r
e
t
r
i
e
v
a
l
o
f
b
r
a
i
n
t
u
m
o
r
i
n
c
o
n
t
r
a
s
t
‐
e
n
h
a
n
c
e
d
M
R
I
i
ma
g
e
s
u
si
n
g
t
u
m
o
r
m
a
r
g
i
n
i
n
f
o
r
ma
t
i
o
n
a
n
d
l
e
a
r
n
e
d
d
i
s
t
a
n
c
e
m
e
t
r
i
c
,
”
M
e
d
i
c
a
l
P
h
y
si
c
s
,
v
o
l
.
3
9
,
n
o
.
1
1
,
p
p
.
6
9
2
9
–
6
9
4
2
,
O
c
t
.
2
0
1
2
,
d
o
i
:
1
0
.
1
1
1
8
/
1
.
4
7
5
4
3
0
5
.
[
2
4
]
A
.
B
e
h
i
n
,
K
.
H
o
a
n
g
-
X
u
a
n
,
A
.
F
.
C
a
r
p
e
n
t
i
e
r
,
a
n
d
J.
-
Y
.
D
e
l
a
t
t
r
e
,
“
P
r
i
mar
y
b
r
a
i
n
t
u
m
o
u
r
s
i
n
a
d
u
l
t
s
,
”
T
h
e
L
a
n
c
e
t
,
v
o
l
.
3
6
1
,
n
o
.
9
3
5
4
,
p
p
.
3
2
3
–
3
3
1
,
J
a
n
.
2
0
0
3
,
d
o
i
:
1
0
.
1
0
1
6
/
s0
1
4
0
-
6
7
3
6
(
0
3
)
1
2
3
2
8
-
8.
[
2
5
]
A
.
D
r
e
v
e
l
e
g
a
s
,
“
I
mag
i
n
g
mo
d
a
l
i
t
i
e
s i
n
b
r
a
i
n
t
u
mo
r
s
,
”
i
n
I
m
a
g
i
n
g
o
f
Br
a
i
n
T
u
m
o
rs
w
i
t
h
H
i
st
o
l
o
g
i
c
a
l
C
o
rre
l
a
t
i
o
n
s
,
S
p
r
i
n
g
e
r
B
e
r
l
i
n
H
e
i
d
e
l
b
e
r
g
,
2
0
0
2
,
p
p
.
1
1
–
2
5
.
[
2
6
]
G
.
Li
t
j
e
n
s
e
t
a
l
.
,
“
A
s
u
r
v
e
y
o
n
d
e
e
p
l
e
a
r
n
i
n
g
i
n
me
d
i
c
a
l
i
m
a
g
e
a
n
a
l
y
si
s
,
”
Me
d
i
c
a
l
I
m
a
g
e
An
a
l
y
si
s
,
v
o
l
.
4
2
,
p
p
.
6
0
–
8
8
,
D
e
c
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
m
e
d
i
a
.
2
0
1
7
.
0
7
.
0
0
5
.
[
2
7
]
Y
.
Le
c
u
n
,
Y
.
B
e
n
g
i
o
,
a
n
d
G
.
H
i
n
t
o
n
,
“
D
e
e
p
l
e
a
r
n
i
n
g
,
”
N
a
t
u
re
,
v
o
l
.
5
2
1
,
n
o
.
7
5
5
3
,
p
p
.
4
3
6
–
4
4
4
,
M
a
y
2
0
1
5
,
d
o
i
:
1
0
.
1
0
3
8
/
n
a
t
u
r
e
1
4
5
3
9
.
[
2
8
]
C
.
P
e
n
g
,
Y
.
Li
u
,
X
.
Y
u
a
n
,
a
n
d
Q
.
C
h
e
n
,
“
R
e
s
e
a
r
c
h
o
f
i
m
a
g
e
r
e
c
o
g
n
i
t
i
o
n
me
t
h
o
d
b
a
se
d
o
n
e
n
h
a
n
c
e
d
i
n
c
e
p
t
i
o
n
-
R
e
sN
e
t
-
V
2
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
1
,
n
o
.
2
4
,
p
p
.
3
4
3
4
5
–
3
4
3
6
5
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
4
2
-
0
2
2
-
1
2
3
8
7
-
0.
[
2
9
]
Y
.
Le
C
u
n
,
L.
B
o
t
t
o
u
,
Y
.
B
e
n
g
i
o
,
a
n
d
P
.
H
a
f
f
n
e
r
,
“
G
r
a
d
i
e
n
t
-
b
a
s
e
d
l
e
a
r
n
i
n
g
a
p
p
l
i
e
d
t
o
d
o
c
u
m
e
n
t
r
e
c
o
g
n
i
t
i
o
n
,
”
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EEE
,
v
o
l
.
8
6
,
n
o
.
1
1
,
p
p
.
2
2
7
8
–
2
3
2
3
,
1
9
9
8
,
d
o
i
:
1
0
.
1
1
0
9
/
5
.
7
2
6
7
9
1
.
[
3
0
]
K.
C
h
i
r
a
n
j
e
e
v
i
,
C.
G
o
m
e
s,
J
.
M
o
r
e
i
r
a
,
L.
S
.
L
o
p
e
s,
V.
C
h
k
o
n
i
y
a
,
A
.
D
.
S
a
n
t
o
s,
a
n
d
C.
M
a
r
q
u
e
s,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
f
o
r
c
a
r
g
o
c
o
d
e
p
r
e
d
i
c
t
i
o
n
i
n
p
o
r
t
l
o
g
i
s
t
i
c
s
,
”
I
n
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
rk
e
t
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
i
e
s
,
S
i
n
g
a
p
o
r
e
:
S
p
r
i
n
g
e
r
N
a
t
u
r
e
S
i
n
g
a
p
o
r
e
,
p
p
.
1
1
5
-
129
,
D
e
c
e
mb
e
r
2
0
2
4
.
[3
1
]
R
.
P
i
l
l
a
i
,
A
.
S
h
a
r
m
a
,
N
.
S
h
a
r
m
a
,
a
n
d
R
.
G
u
p
t
a
,
“
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
V
G
G
1
6
,
R
e
sN
e
t
5
0
,
a
n
d
i
n
c
e
p
t
i
o
n
V
3
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
m
o
d
e
l
s
,
”
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
N
O
C
O
N
5
7
9
7
5
.
2
0
2
3
.
1
0
1
0
1
2
5
2
.
[3
2
]
A
.
S
a
l
e
h
,
R
.
S
u
k
a
i
k
,
a
n
d
S
.
S
.
A
b
u
-
N
a
ser,
“
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
d
e
e
p
l
e
a
r
n
i
n
g
,
”
i
n
2
0
2
0
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Assi
st
i
v
e
a
n
d
Re
h
a
b
i
l
i
t
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s (i
C
a
r
e
T
e
c
h
)
,
A
u
g
.
2
0
2
0
,
p
p
.
1
3
1
–
1
3
6
,
d
o
i
:
1
0
.
1
1
0
9
/
i
c
a
r
e
t
e
c
h
4
9
9
1
4
.
2
0
2
0
.
0
0
0
3
2
.
[3
3
]
Y
.
Y
a
n
g
,
C
h
a
o
l
u
o
m
e
n
g
,
a
n
d
N
.
R
a
z
mj
o
o
y
,
“
Ea
r
l
y
d
e
t
e
c
t
i
o
n
o
f
b
r
a
i
n
t
u
m
o
r
s:
H
a
r
n
e
ss
i
n
g
t
h
e
p
o
w
e
r
o
f
G
R
U
n
e
t
w
o
r
k
s
a
n
d
h
y
b
r
i
d
d
w
a
r
f
m
o
n
g
o
o
s
e
o
p
t
i
mi
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
,
”
Bi
o
m
e
d
i
c
a
l
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
9
1
,
p
.
1
0
6
0
9
3
,
M
a
y
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
s
p
c
.
2
0
2
4
.
1
0
6
0
9
3
.
[3
4
]
D
.
R
e
y
e
s
a
n
d
J.
S
á
n
c
h
e
z
,
“
P
e
r
f
o
r
m
a
n
c
e
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
t
h
e
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
b
r
a
i
n
t
u
m
o
r
s
u
s
i
n
g
ma
g
n
e
t
i
c
r
e
so
n
a
n
c
e
i
m
a
g
i
n
g
,
”
H
e
l
i
y
o
n
,
v
o
l
.
1
0
,
n
o
.
3
,
p
.
e
2
5
4
6
8
,
F
e
b
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
h
e
l
i
y
o
n
.
2
0
2
4
.
e
2
5
4
6
8
.
[3
5
]
P
.
P
o
d
d
e
r
,
S
.
B
h
a
r
a
t
i
,
M
.
A
.
R
a
h
ma
n
,
a
n
d
U
.
K
o
se,
“
Tr
a
n
sf
e
r
l
e
a
r
n
i
n
g
f
o
r
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
b
r
a
i
n
t
u
m
o
r
,
”
i
n
D
e
e
p
L
e
a
rn
i
n
g
f
o
r
Bi
o
m
e
d
i
c
a
l
A
p
p
l
i
c
a
t
i
o
n
s
,
C
R
C
P
r
e
ss
,
2
0
2
1
,
p
p
.
3
1
5
–
3
2
8
.
[3
6
]
P
.
G
h
o
s
a
l
,
L.
N
a
n
d
a
n
w
a
r
,
S
.
K
a
n
c
h
a
n
,
A
.
B
h
a
d
r
a
,
J
.
C
h
a
k
r
a
b
o
r
t
y
,
a
n
d
D
.
N
a
n
d
i
,
“
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
R
e
sN
e
t
-
1
0
1
b
a
s
e
d
sq
u
e
e
z
e
a
n
d
e
x
c
i
t
a
t
i
o
n
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
i
n
2
0
1
9
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
A
d
v
a
n
c
e
d
C
o
m
p
u
t
a
t
i
o
n
a
l
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
Pa
r
a
d
i
g
m
s (I
C
A
C
C
P
)
,
F
e
b
.
2
0
1
9
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
i
c
a
c
c
p
.
2
0
1
9
.
8
8
8
2
9
7
3
.
[3
7
]
“
B
r
a
i
n
t
u
m
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
(
M
R
I
)
.
”
h
t
t
p
s:
/
/
w
w
w
.
k
a
g
g
l
e
.
c
o
m/
d
a
t
a
s
e
t
s
/
sart
a
j
b
h
u
v
a
j
i
/
b
r
a
i
n
-
t
u
mo
r
-
c
l
a
ss
i
f
i
c
a
t
i
o
n
-
mr
i
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
K
a
lp
a
n
a
Te
lk
a
r
re
c
e
iv
e
d
th
e
B.
Tec
h
.
d
e
g
re
e
in
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
in
2
0
0
2
a
n
d
M
.
Tec
h
.
d
e
g
re
e
in
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
fr
o
m
Ja
wa
h
a
rlal
Tec
h
n
o
l
o
g
ica
l
Un
i
v
e
rsity
Ka
k
i
n
a
d
a
,
An
d
h
ra
P
ra
d
e
sh
,
I
n
d
ia i
n
2
0
1
1
.
At
p
re
se
n
t
sh
e
is p
u
rsu
i
n
g
P
h
.
D.
a
t
t
h
e
De
p
a
rtme
n
t
of
El
e
c
t
ro
n
ics
E
n
g
in
e
e
rin
g
,
S
c
h
o
o
l
o
f
E
n
g
in
e
e
rin
g
a
n
d
Tec
h
n
o
lo
g
y
,
P
o
n
d
ich
e
rry
U
n
iv
e
rsit
y
,
P
o
n
d
ich
e
rry
,
In
d
ia.
He
r
re
se
a
rc
h
in
tere
st
in
c
lu
d
e
ML
,
DL
,
a
n
d
ima
g
e
p
ro
c
e
ss
in
g
.
S
h
e
is
t
h
e
m
e
m
b
e
r
o
f
IE
TE
a
n
d
I
S
TE
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
telk
a
rk
a
lp
a
n
a
@g
m
a
il
.
c
om
.
Dr
.
K
.
Anu
sud
h
a
re
c
e
iv
e
d
t
h
e
B.
E
d
e
g
re
e
i
n
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
En
g
i
n
e
e
rin
g
fr
o
m
M
a
d
ra
s
Un
i
v
e
rsity
,
M
.
E
d
e
g
re
e
in
Co
m
m
u
n
i
c
a
ti
o
n
S
y
ste
m
s
fr
o
m
An
n
a
Un
iv
e
rsity
,
P
h
.
D
.
d
e
g
re
e
in
El
e
c
tro
n
ics
En
g
in
e
e
rin
g
/Alg
o
rit
h
m
s
fo
r
S
e
c
u
re
Tran
sa
c
ti
o
n
o
f
M
e
d
ica
l
Im
a
g
e
s
fro
m
VIT
U
n
i
v
e
rsity
a
n
d
M
BA
in
Hu
m
a
n
R
e
so
u
rc
e
M
a
n
a
g
e
m
e
n
t
fro
m
P
o
n
d
ich
e
rry
U
n
iv
e
rsit
y
.
S
h
e
is
c
u
rre
n
tl
y
a
n
As
so
c
iate
P
ro
fe
ss
o
r
i
n
De
p
a
rtme
n
t
o
f
El
e
c
tro
n
ics
En
g
i
n
e
e
rin
g
,
S
c
h
o
o
l
o
f
En
g
in
e
e
r
in
g
a
n
d
Tec
h
n
o
l
o
g
y
,
P
o
n
d
ic
h
e
rry
Un
iv
e
rsity
,
P
o
n
d
ich
e
rr
y
,
In
d
ia.
He
r
re
se
a
rc
h
i
n
tere
sts
in
c
lu
d
e
d
ig
i
tal
si
g
n
a
l
a
n
d
ima
g
e
p
r
o
c
e
ss
in
g
,
a
n
d
a
rti
ficia
l
in
telli
g
e
n
c
e
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
n
u
su
d
h
a
k
@
y
a
h
o
o
.
c
o
.
i
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.