I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
10
,
No
.
3
,
Sep
tem
b
er
202
1
,
p
p
.
5
7
6
~
5
8
3
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
10
.i
3
.
p
p
5
7
6
-
5
8
3
576
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Autom
a
tic
c
erebr
o
v
a
scula
r seg
menta
tion met
ho
ds
-
a
revi
ew
F
a
t
ma
T
a
her,
Nee
m
a
P
r
a
ka
s
h
Co
ll
e
g
e
o
f
Tec
h
n
o
lo
g
ica
l
In
n
o
v
a
t
io
n
,
Zay
e
d
U
n
iv
e
rsit
y
,
D
u
b
a
i,
Un
it
e
d
Ara
b
Emira
tes
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
2
6
,
2
0
2
0
R
ev
is
ed
Ma
y
1
8
,
2
0
2
1
Acc
ep
ted
Ma
y
2
7
,
2
0
2
1
Ce
re
b
ro
v
a
sc
u
lar
d
ise
a
se
s
a
re
o
n
e
o
f
th
e
se
ri
o
u
s
c
a
u
se
s
fo
r
th
e
in
c
re
a
se
in
m
o
rtalit
y
ra
te
i
n
t
h
e
wo
rl
d
w
h
ich
a
ffe
c
t
th
e
b
l
o
o
d
v
e
ss
e
ls
a
n
d
b
lo
o
d
su
p
p
l
y
t
o
th
e
b
ra
in
.
In
o
rd
e
r,
d
iag
n
o
s
e
a
n
d
stu
d
y
t
h
e
a
b
n
o
rm
a
li
ti
e
s
in
th
e
c
e
re
b
ro
v
a
sc
u
lar
sy
ste
m
,
a
c
c
u
ra
t
e
se
g
m
e
n
tatio
n
m
e
th
o
d
s
c
a
n
b
e
u
se
d
.
T
h
e
sh
a
p
e
,
d
irec
ti
o
n
a
n
d
d
istri
b
u
t
io
n
o
f
b
lo
o
d
v
e
ss
e
ls
c
a
n
b
e
stu
d
ied
u
si
n
g
a
u
to
m
a
ti
c
se
g
m
e
n
tatio
n
.
T
h
is
will
h
e
lp
th
e
d
o
c
to
rs
to
e
n
v
isa
g
e
th
e
c
e
re
b
ro
v
a
sc
u
lar
sy
ste
m
.
D
u
e
to
th
e
c
o
m
p
le
x
sh
a
p
e
a
n
d
t
o
p
o
lo
g
y
,
a
u
t
o
m
a
ti
c
se
g
m
e
n
tatio
n
is
stil
l
a
c
h
a
ll
e
n
g
e
to
th
e
c
li
n
icia
n
s.
In
t
h
is
p
a
p
e
r,
s
o
m
e
o
f
th
e
late
st
a
p
p
ro
a
c
h
e
s
u
se
d
f
o
r
se
g
m
e
n
tatio
n
o
f
m
a
g
n
e
ti
c
re
so
n
a
n
c
e
a
n
g
io
g
ra
p
h
y
ima
g
e
s
a
re
e
x
p
lain
e
d
.
S
o
m
e
o
f
s
u
c
h
m
e
th
o
d
s
a
re
d
e
e
p
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
(CNN
),
3
d
ime
n
ti
o
n
a
l
-
C
NN
(3
D
-
CNN
)
a
n
d
3
D
U
-
Ne
t.
F
i
n
a
ll
y
,
t
h
e
se
m
e
th
o
d
s
a
re
c
o
m
p
a
re
d
fo
r
e
v
a
l
u
a
ti
n
g
t
h
e
ir
p
e
rfo
rm
a
n
c
e
.
3
D
U
-
Ne
t
is
t
h
e
b
e
tt
e
r
p
e
rfo
rm
e
r
a
m
o
n
g
t
h
e
d
e
sc
ri
b
e
d
m
e
th
o
d
s.
K
ey
w
o
r
d
s
:
C
er
eb
r
o
v
ascu
lar
C
NN
MRA
Seg
m
en
tatio
n
U
-
Net
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
:
Fatm
a
T
ah
er
C
o
lleg
e
o
f
T
ec
h
n
o
lo
g
ical
I
n
n
o
v
atio
n
Z
ay
ed
Un
iv
e
r
s
ity
Du
b
ai,
Un
ited
Ar
ab
E
m
ir
ates
E
m
ail: f
atm
a.
tah
er
@
zu
.
ac
.
ae
1.
I
NT
RO
D
UCT
I
O
N
I
n
th
e
w
o
r
ld
,
v
ascu
lar
d
is
ea
s
es
ar
e
o
n
e
o
f
t
h
e
m
o
s
t
co
m
m
o
n
ca
u
s
e
o
f
d
ea
th
wh
ich
ca
u
s
es
s
tr
o
k
e
in
ab
o
u
t
m
illi
o
n
s
o
f
p
eo
p
le
ev
er
y
y
ea
r
.
T
h
er
e
f
o
r
e,
f
ast,
an
d
ac
cu
r
ate
to
o
ls
ar
e
r
eq
u
ir
ed
in
o
r
d
er
to
d
iag
n
o
s
e
an
d
tr
ea
t
ce
r
eb
r
o
v
ascu
lar
d
is
ea
s
es.
Ma
g
n
etic
r
eso
n
a
n
ce
an
g
io
g
r
ap
h
y
(
MRA
)
is
th
e
o
n
e
o
f
th
e
co
m
m
o
n
im
ag
in
g
tech
n
iq
u
es
u
s
ed
t
o
p
er
f
o
r
m
th
is
f
u
n
ctio
n
,
w
h
ich
co
n
s
is
ts
in
a
m
ag
n
etic
r
eso
n
an
ce
im
a
g
in
g
(
MRI)
th
at
lo
o
k
s
s
p
ec
if
ically
th
e
b
lo
o
d
f
lo
w
i
n
th
e
b
r
ain
v
ess
els
wh
en
m
e
asu
r
in
g
.
Dif
f
e
r
en
t
m
et
h
o
d
s
o
f
MRA
ar
e
ti
me
-
of
-
f
lig
h
t
(
T
OF)
,
p
h
ase
co
n
tr
ast
(
PC
)
,
an
d
f
r
esh
b
l
o
o
d
im
a
g
i
n
g
(
FB
I
)
an
d
c
o
n
tr
ast
-
en
h
an
c
ed
MRA
[
1
]
.
T
OF
MRA
i
s
th
e
m
o
s
t
co
m
m
o
n
ly
u
s
ed
im
ag
in
g
m
o
d
alities
in
n
o
n
-
in
v
asiv
e
v
ascu
lar
r
esear
ch
[
2
]
.
Seg
m
en
tatio
n
is
u
s
ed
to
id
en
tif
y
an
d
s
ep
ar
ate
v
ess
els
f
r
o
m
n
eig
h
b
o
r
h
o
o
d
tis
s
u
e
wh
ich
h
elp
s
in
b
etter
v
ie
w
an
d
q
u
an
titativ
e
an
aly
s
is
.
Seg
m
en
tatio
n
o
f
b
l
o
o
d
v
ess
els
d
o
n
e
m
an
u
ally
is
h
av
in
g
m
an
y
s
h
o
r
tc
o
m
in
g
s
lik
e
it
is
tim
e
-
co
n
s
u
m
in
g
,
p
r
o
n
e
t
o
er
r
o
r
.
I
n
s
u
ch
a
s
itu
atio
n
,
m
o
r
e
a
cc
u
r
ate
an
d
f
aster
s
eg
m
en
ta
tio
n
m
eth
o
d
s
ar
e
im
p
lem
en
ted
.
So
m
e
o
f
th
e
ap
p
r
o
ac
h
es
u
s
e
d
f
o
r
ce
r
eb
r
o
v
as
cu
lar
s
eg
m
en
tatio
n
ar
e
d
ee
p
C
NN
[
3
]
,
3
D
C
NN,
an
d
3
D
U
-
Net.
A
C
NN
is
a
class
o
f
d
ee
p
n
eu
r
al
n
etwo
r
k
[
4
]
c
o
n
s
is
tin
g
o
f
o
n
e
o
r
m
o
r
e
c
o
n
v
o
lu
tio
n
al
lay
e
r
s
,
p
o
o
lin
g
lay
er
,
an
d
f
u
lly
c
o
n
n
ec
ted
lay
er
.
A
f
ea
tu
r
e
m
ap
cr
ea
ted
b
y
th
e
f
ir
s
t
lay
er
i.e
.
,
co
n
v
o
lu
tio
n
al
lay
er
is
u
s
ed
to
ex
tr
ac
t th
e
f
ea
tu
r
es f
r
o
m
an
in
p
u
t im
ag
e.
T
h
is
is
p
o
s
s
ib
le
b
y
u
s
in
g
a
f
ilter
th
at
s
ca
n
s
th
e
f
u
ll im
ag
e
p
ix
el
wis
e
.
Po
o
lin
g
la
y
er
c
u
td
o
wn
th
e
q
u
an
tity
o
f
in
f
o
r
m
atio
n
th
e
f
ir
s
t
lay
er
g
e
n
er
ated
f
o
r
ea
c
h
f
ea
tu
r
e
an
d
m
ai
n
tain
s
th
e
m
o
s
t
im
p
o
r
tan
t
in
f
o
r
m
atio
n
o
n
ly
.
T
h
e
o
u
t
p
u
t
g
e
n
er
ated
b
y
th
e
p
o
o
lin
g
lay
e
r
is
f
latten
ed
b
y
f
u
lly
co
n
n
ec
ted
in
p
u
t
lay
er
,
wh
ich
co
n
v
er
ts
th
em
in
to
a
s
in
g
le
v
ec
to
r
th
at
s
e
r
v
es
as
an
in
p
u
t
f
o
r
t
h
e
n
e
x
t
l
ay
er
.
Af
ter
p
ass
in
g
th
r
o
u
g
h
th
e
f
u
lly
co
n
n
ec
ted
lay
er
,
t
h
e
f
in
al
lay
e
r
s
u
s
e
th
e
So
f
tMa
x
ac
tiv
atio
n
f
u
n
ctio
n
wh
ich
h
elp
s
in
class
if
icatio
n
.
Fu
lly
co
n
n
ec
te
d
o
u
t
p
u
t
lay
e
r
g
e
n
er
ates
th
e
f
in
al
r
esu
lt
wh
ich
will
d
eter
m
in
e
a
class
f
o
r
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
ma
tic
ce
r
eb
r
o
va
s
cu
la
r
s
eg
men
ta
tio
n
meth
o
d
s
-
a
r
ev
iew
(
F
a
tma
Ta
h
er
)
577
im
ag
e.
T
h
e
f
ilter
s
o
f
d
if
f
er
e
n
t
s
izes
ar
e
u
s
ed
,
al
s
o
k
n
o
wn
as
k
er
n
els.
Usu
ally
,
co
n
v
o
lu
ti
o
n
lay
er
an
d
p
o
o
l
lay
er
ar
e
u
s
ed
in
s
o
m
e
c
o
m
b
in
atio
n
in
C
NN
ar
ch
itectu
r
e
[
5
]
.
Ma
x
p
o
o
lin
g
an
d
m
ea
n
p
o
o
lin
g
a
r
e
th
e
two
ty
p
es
o
f
o
p
er
atio
n
s
ca
r
r
ie
d
o
u
t
b
y
th
e
p
o
o
lin
g
lay
e
r
.
Me
an
p
o
o
lin
g
o
r
av
er
ag
e
p
o
o
lin
g
ca
lcu
lates
th
e
av
er
ag
e
v
alu
e
f
r
o
m
th
e
r
eg
i
o
n
o
f
th
e
f
ea
tu
r
e
m
ap
.
Ma
x
im
u
m
Po
o
lin
g
also
ca
lled
as
m
ax
p
o
o
li
n
g
f
in
d
s
th
e
m
ax
im
u
m
v
alu
e
f
r
o
m
th
e
r
e
g
io
n
o
f
f
ea
tu
r
e
m
ap
co
v
er
e
d
b
y
f
ilter
.
T
h
e
er
r
o
r
ca
u
s
ed
b
y
th
e
n
eig
h
b
o
r
h
o
o
d
s
ize
lim
itatio
n
ca
n
b
e
r
ed
u
ce
d
b
y
m
ea
n
p
o
o
li
n
g
an
d
it
also
r
etain
s
th
e
b
ac
k
g
r
o
u
n
d
in
f
o
r
m
atio
n
.
T
h
e
esti
m
ated
er
r
o
r
ca
u
s
ed
b
y
th
e
m
ea
n
d
e
v
iatio
n
ca
n
b
e
r
ed
u
ce
d
b
y
m
a
x
p
o
o
lin
g
a
n
d
h
e
n
ce
k
ee
p
s
m
o
r
e
tex
tu
r
e
in
f
o
r
m
atio
n
.
I
n
3
D
-
C
NN,
3
D
f
ilter
s
ca
n
m
o
v
e
in
all
th
e
th
r
ee
d
im
e
n
s
io
n
s
i.e
.
,
alo
n
g
X,
Y
a
n
d
Z
ax
es.
3
D
C
NN
r
eq
u
ir
es
m
o
r
e
p
ar
am
eter
s
an
d
co
m
p
u
tatio
n
s
co
m
p
ar
ed
to
2
D.
I
m
p
r
o
v
em
e
n
t
in
c
o
m
p
u
ter
h
ar
d
war
e
an
d
3
D
m
ed
ical
im
ag
in
g
av
aila
b
ilit
y
g
iv
es
th
e
co
n
ce
p
t
o
f
u
s
in
g
3
D
in
f
o
r
m
atio
n
f
o
r
s
eg
m
e
n
tatio
n
.
C
o
m
p
ar
e
d
to
2
D
an
d
2
.
5
D
ap
p
r
o
ac
h
es
wh
ich
a
r
e
h
av
in
g
o
n
e
,
th
r
ee
o
r
th
o
g
o
n
al
v
iews,
r
esp
ec
tiv
ely
,
3
d
im
e
n
s
io
n
al
im
ag
es
ca
n
d
eliv
er
in
f
o
r
m
atio
n
in
an
y
d
ir
ec
tio
n
.
Fo
r
th
e
s
eg
m
e
n
tatio
n
o
f
th
e
b
r
ain
tu
m
o
r
o
f
a
r
b
itra
r
y
s
ize,
th
e
f
ir
s
t
p
u
r
e
3
D
m
o
d
els
wer
e
in
tr
o
d
u
ce
d
.
Mu
ltis
ca
le,
d
u
al
-
p
ath
3
D
C
NN
is
also
u
s
ed
in
m
an
y
ap
p
licatio
n
s
.
T
h
e
n
ex
t
p
ath
way
r
ec
eiv
ed
th
e
p
atch
es
f
r
o
m
a
s
u
b
s
am
p
le
d
r
ep
r
esen
tatio
n
o
f
th
e
im
a
g
e.
Mo
r
e
a
r
ea
s
ar
o
u
n
d
th
e
v
o
x
el
ca
n
b
e
p
r
o
ce
s
s
ed
b
y
u
s
in
g
th
i
s
,
wh
ich
will
b
e
a
d
v
an
ta
g
eo
u
s
to
th
e
wh
o
le
s
y
s
tem
[
6
]
.
Fo
r
b
etter
p
er
f
o
r
m
a
n
ce
,
s
m
all
k
er
n
el
s
ize
ca
n
b
e
p
r
ef
e
r
r
ed
.
A
d
ee
p
m
o
d
el
is
r
eq
u
ir
e
d
to
is
o
late
a
n
o
r
g
an
f
r
o
m
co
m
p
lex
im
ag
es
w
h
ich
ca
n
th
er
eb
y
ex
tr
ac
t
h
ig
h
ly
i
n
f
o
r
m
ativ
e
f
ea
tu
r
es.
A
s
ig
n
if
ica
n
t
ch
allen
g
e
f
o
r
3
D
m
o
d
els
is
to
tr
ain
s
u
ch
d
ee
p
n
etwo
r
k
.
I
n
o
r
d
er
to
s
tr
en
g
th
en
th
e
U
-
Net
[
7
]
s
tr
u
ct
u
r
e
with
r
ich
er
s
p
atial
in
f
o
r
m
atio
n
m
o
d
el,
3
D
U
-
Net
is
d
ev
elo
p
e
d
wh
ich
is
a
d
ee
p
n
eu
r
al
n
etwo
r
k
th
at
h
elp
s
in
v
er
y
co
m
p
ac
t
v
o
lu
m
etr
ic
s
eg
m
en
tatio
n
.
3
D
U
-
Net
r
eq
u
ir
es
o
n
ly
s
o
m
e
a
n
n
o
tated
2
D
s
lices
b
y
u
s
in
g
weig
h
ted
lo
s
s
f
u
n
ctio
n
an
d
d
ata
au
g
m
en
tatio
n
f
o
r
tr
ain
in
g
.
T
h
is
n
etwo
r
k
tak
es
3
D
v
o
lu
m
e
as
in
p
u
t
an
d
3
D
o
p
er
atio
n
s
l
ik
e
co
n
v
o
l
u
tio
n
,
p
o
o
lin
g
,
a
n
d
lo
s
s
ca
lcu
latio
n
ar
e
u
s
ed
to
p
r
o
ce
s
s
th
em
.
Fo
r
v
ascu
lar
b
o
u
n
d
ar
y
d
etec
tio
n
,
3
D
U
-
Net
was
u
s
ed
.
On
e
o
f
th
e
d
is
ad
v
an
tag
es
o
f
3
D
U
-
Net
is
th
e
in
p
u
t
im
a
g
e
s
ize
s
h
o
u
ld
b
e
s
m
all
b
ec
au
s
e
o
f
li
m
ited
m
em
o
r
y
s
p
ac
e
[
8
]
.
T
h
er
ef
o
r
e,
t
h
e
in
p
u
t
s
ize
o
f
r
e
g
io
n
o
f
in
ter
est
(
R
OI
)
is
h
av
in
g
p
o
o
r
r
eso
lu
tio
n
.
T
h
e
r
e
f
o
r
e,
th
e
in
p
u
t
i
m
ag
e
ca
n
b
e
d
iv
id
ed
in
t
o
m
u
ltip
le
b
atch
es
to
o
v
er
co
m
e
th
is
is
s
u
e,
wh
ich
ca
n
b
e
f
u
r
t
h
er
u
s
ed
f
o
r
tr
ain
in
g
an
d
test
in
g
.
I
n
th
is
p
ap
er
,
th
ese
th
r
ee
m
eth
o
d
s
ar
e
ex
p
lain
ed
in
d
etai
l a
n
d
co
m
p
ar
ed
th
ei
r
p
er
f
o
r
m
a
n
ce
with
g
lo
b
al
s
tatis
tica
l b
ase
d
ap
p
r
o
ac
h
(
GSB
)
b
y
u
s
in
g
d
ice
s
im
ilar
ity
co
ef
f
i
cien
t (
DSC
)
v
alu
es.
T
h
e
r
em
ain
in
g
p
ar
t
o
f
th
is
p
a
p
er
is
o
r
g
a
n
ized
is
b
ein
g
as.
I
n
th
e
n
e
x
t
s
ec
tio
n
,
liter
atu
r
e
s
u
r
v
ey
is
p
r
es
en
ted
.
I
n
s
ec
tio
n
3
,
d
if
f
er
en
t
ce
r
eb
r
o
v
ascu
lar
s
eg
m
en
ta
tio
n
m
eth
o
d
s
ar
e
e
x
p
lain
ed
,
i
n
s
ec
tio
n
4
,
r
esu
lts
ar
e
d
is
cu
s
s
ed
an
d
f
in
ally
co
n
cl
u
s
io
n
is
p
r
esen
ted
in
s
ec
tio
n
5
.
2.
RE
L
AT
E
D
WO
RK
S
Ma
jo
r
co
n
tr
ib
u
tio
n
s
o
f
s
o
m
e
o
f
th
e
r
esear
ch
er
s
wh
o
aim
ed
at
d
ev
elo
p
in
g
a
s
y
s
tem
f
o
r
ce
r
eb
r
o
v
ascu
lar
s
eg
m
en
tatio
n
ar
e
s
u
m
m
ar
ize
d
b
elo
w.
San
ch
es
et
a
l.
[
1
]
p
r
o
p
o
s
e
d
a
ce
r
eb
r
o
v
ascu
lar
s
eg
m
en
tatio
n
m
eth
o
d
u
s
in
g
d
ee
p
lear
n
in
g
.
A
3
D
m
o
d
el
ca
lled
Uce
p
tio
n
wh
ich
is
in
s
p
ir
ed
f
r
o
m
U
-
Net
ar
ch
itectu
r
e
is
d
is
cu
s
s
ed
in
th
is
p
ap
er
.
W
h
en
co
m
p
ar
ed
with
th
e
U
-
Net
m
o
d
el,
th
is
3
D
a
r
ch
itectu
r
e
s
h
o
wed
b
etter
p
er
f
o
r
m
a
n
ce
.
I
n
o
r
d
e
r
t
o
im
p
r
o
v
e
th
e
o
u
tco
m
e
o
f
th
i
s
m
o
d
el,
t
h
ey
h
av
e
d
ec
id
e
d
t
o
ad
d
m
o
r
e
d
etails
r
eg
ar
d
in
g
th
e
ce
r
e
b
r
o
v
ascu
lar
an
ato
m
y
in
th
e
n
e
u
r
al
n
etwo
r
k
.
Hesam
ian
et
a
l.
[
6
]
s
u
m
m
a
r
ized
s
o
m
e
o
f
th
e
m
ed
ical
i
m
ag
e
s
eg
m
e
n
tatio
n
m
eth
o
d
s
an
d
th
eir
p
er
f
o
r
m
an
ce
co
m
p
a
r
ed
with
o
ld
m
eth
o
d
s
.
T
h
is
p
ap
e
r
also
e
x
p
lain
ed
s
o
m
e
o
f
th
e
a
p
p
licat
io
n
s
h
elp
f
u
l
in
th
e
m
ed
ical
in
d
u
s
tr
y
s
u
ch
as
tr
ain
in
g
tech
n
iq
u
es
u
s
ed
f
o
r
im
a
g
e
s
eg
m
en
tatio
n
.
Ad
v
a
n
tag
es
an
d
d
is
ad
v
a
n
tag
es
o
f
th
ese
tech
n
iq
u
es
ar
e
also
tak
en
in
to
co
n
s
id
er
atio
n
.
T
h
e
ch
a
llen
g
es
f
ac
ed
b
y
th
e
d
ee
p
lea
r
n
in
g
n
etwo
r
k
s
f
o
r
s
eg
m
en
tatio
n
an
d
its
ef
f
ec
tiv
e
r
em
ed
ies ar
e
also
ex
p
lain
e
d
at
th
e
en
d
.
A
f
u
lly
au
to
m
atic
s
eg
m
en
t
atio
n
m
eth
o
d
is
p
r
o
p
o
s
ed
b
y
Gao
et
a
l.
[
9
]
f
o
r
d
et
ec
tio
n
o
f
ce
r
eb
r
o
v
ascu
lar
d
is
ea
s
es.
T
h
is
s
eg
m
en
tatio
n
m
eth
o
d
is
v
er
y
f
ast
to
o
.
I
m
p
r
o
v
ed
cu
r
v
e
ev
o
l
u
tio
n
an
d
s
tatis
tical
m
o
d
el
an
al
y
s
is
p
lay
a
m
ajo
r
r
o
le
in
th
e
s
eg
m
en
tatio
n
o
f
3
D
-
ce
r
eb
r
al
v
ess
els
f
r
o
m
MRA.
Mo
d
ellin
g
o
f
th
e
ce
r
eb
r
al
v
ess
els
is
al
s
o
ex
p
lain
ed
in
th
is
p
ap
er
.
C
o
m
b
in
atio
n
o
f
r
e
g
io
n
d
is
tr
ib
u
tio
n
an
d
g
r
ad
ien
t
in
f
o
r
m
atio
n
is
u
s
ed
as
o
n
e
n
o
v
el
m
o
d
e
i
n
cu
r
v
e
ev
o
lu
tio
n
.
L
o
w
c
o
n
tr
ast
th
in
v
ess
el
b
o
u
n
d
ar
y
a
r
o
u
n
d
b
r
ain
tis
s
u
e
ca
n
b
e
d
eter
m
in
ed
b
y
u
s
in
g
t
h
e
ed
g
e
-
s
tr
en
g
th
f
u
n
ctio
n
.
A
f
ast
lev
el
s
et
m
eth
o
d
was
in
tr
o
d
u
c
ed
to
s
p
ee
d
u
p
th
e
im
p
lem
en
tatio
n
o
f
cu
r
v
e
e
v
o
lu
tio
n
wh
ic
h
h
elp
s
in
i
m
p
r
o
v
i
n
g
th
e
p
er
f
o
r
m
an
ce
o
f
ce
r
e
b
r
o
v
ascu
lar
s
eg
m
en
tatio
n
.
Fatm
a
et
a
l.
[
1
0
]
p
r
esen
te
d
a
r
ev
iew
o
n
ac
c
u
r
ate
a
n
d
ad
v
an
ce
d
au
to
m
ated
m
e
th
o
d
s
f
o
r
ce
r
eb
r
o
v
ascu
lar
s
eg
m
e
n
tatio
n
.
I
n
th
is
p
a
p
er
,
o
ld
,
n
ew,
au
to
m
atic,
an
d
s
em
iau
to
m
atic
m
o
d
els
ex
p
lain
ed
alo
n
g
with
its
ad
v
an
tag
es
an
d
d
is
ad
v
an
tag
es.
A
lin
ea
r
co
m
b
i
n
atio
n
o
f
d
is
cr
ete
g
au
s
s
ian
s
(
L
C
DG)
m
o
d
el
is
u
s
ed
f
o
r
s
eg
m
en
tatio
n
th
at
y
ield
s
th
e
em
p
ir
ical
m
ar
g
in
al
g
r
ay
lev
el
d
is
tr
ib
u
ti
o
n
in
ten
s
ity
in
th
e
i
m
ag
es,
wh
ile
u
s
in
g
m
o
d
if
ied
e
x
p
ec
tatio
n
m
ax
im
iz
atio
n
(
E
M)
alg
o
r
ith
m
f
o
r
r
ef
in
em
en
t.
A
s
tatis
tical
m
eth
o
d
d
is
cu
s
s
ed
b
y
Fatm
a
et
a
l.
[
1
1
]
u
tili
ze
s
a
v
o
x
el
-
wis
e
class
if
icatio
n
.
I
n
o
r
d
er
to
is
o
late
b
lo
o
d
v
ess
els
f
r
o
m
th
e
b
ac
k
g
r
o
u
n
d
o
f
ea
ch
tim
e
o
f
f
lig
h
t
MRA
s
lice,
p
r
o
b
ab
ilit
y
m
o
d
els
o
f
v
o
x
el
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
10
,
N
o
.
3
,
Sep
tem
b
er
202
1
:
5
7
6
-
583
578
in
ten
s
ities
ar
e
d
eter
m
in
ed
.
T
h
e
m
ar
g
in
al
em
p
ir
ical
d
is
tr
ib
u
t
io
n
o
f
in
ten
s
ity
p
r
o
b
ab
ilit
ies
is
ap
p
r
o
x
im
ated
f
o
r
th
e
p
u
r
p
o
s
e
o
f
class
if
icatio
n
wh
er
e
L
C
DG
is
em
p
lo
y
ed
with
alter
n
ate
s
ig
n
s
.
Fo
r
lin
ea
r
co
m
b
in
atio
n
o
f
g
au
s
s
ian
ap
p
r
o
x
im
atio
n
,
E
M
-
b
ased
tech
n
iq
u
es a
r
e
also
u
tili
ze
d
th
at
h
elp
s
in
d
ea
lin
g
with
L
C
DGs.
3.
RE
S
E
ARCH
M
E
T
H
O
D
C
er
eb
r
o
v
ascu
lar
s
eg
m
en
tatio
n
m
eth
o
d
s
s
u
ch
as
d
ee
p
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
,
3
d
im
en
tio
n
al
-
C
NN
(
3
D
-
C
NN)
an
d
3
D
U
-
Net
ar
e
u
s
ed
f
o
r
co
n
d
u
ctin
g
th
is
r
esear
ch
wh
ic
h
ar
e
d
is
cu
s
s
ed
i
n
d
etail.
T
h
ese
m
eth
o
d
s
ar
e
co
m
p
ar
ed
f
o
r
ev
alu
atin
g
th
eir
p
er
f
o
r
m
a
n
ce
s
.
Dice
s
im
ilar
ity
co
ef
f
icien
t
(
DSC
)
is
u
s
ed
f
o
r
d
eter
m
in
in
g
th
e
s
eg
m
en
tatio
n
ac
cu
r
ac
y
.
3
.
1
.
Co
nv
o
lutio
na
l neura
l net
wo
rk
(
CNN)
W
h
en
co
n
s
id
er
in
g
tim
e
-
of
-
f
li
g
h
t
(
T
OF
)
MRA
im
ag
es,
th
er
e
ar
e
ch
an
ce
s
o
f
o
v
e
r
-
f
itti
n
g
in
t
h
e
m
o
d
el
lear
n
in
g
an
d
in
cr
ea
s
ed
p
r
o
ce
s
s
in
g
tim
e
b
y
th
e
u
s
e
o
f
c
o
m
p
l
ex
d
ee
p
C
NN
ar
ch
itectu
r
es.
A
C
NN
ar
ch
itectu
r
e
p
r
o
p
o
s
ed
b
y
Ph
ellan
et
a
l.
[
1
2
]
c
o
m
p
o
s
ed
o
f
two
co
n
v
o
lu
t
io
n
al
lay
er
s
an
d
f
u
lly
co
n
n
ec
t
ed
lay
er
s
.
Fig
u
r
e
1
d
ep
icts
th
e
C
NN
ar
ch
itectu
r
e,
wh
er
e
th
e
n
u
m
b
er
o
f
f
ilter
s
an
d
d
etails
o
f
r
ec
ep
tiv
e
f
ield
in
th
e
two
co
n
v
o
l
u
tio
n
al
lay
er
s
ar
e
m
en
ti
o
n
ed
.
T
h
er
e
is
n
o
s
u
b
s
am
p
lin
g
in
th
e
s
ec
o
n
d
c
o
n
v
o
lu
tio
n
al
lay
er
.
Nex
t
lay
e
r
is
a
r
ec
tifie
d
lin
ea
r
ac
tiv
atio
n
(
R
eL
u
)
wh
ich
will
r
ed
u
ce
th
e
b
a
ck
p
r
o
p
a
g
atio
n
v
a
n
is
h
in
g
p
r
o
b
lem
.
T
h
er
e
ar
e
two
f
u
lly
co
n
n
ec
te
d
lay
er
s
p
r
esen
t
in
th
is
ar
ch
itectu
r
e.
T
h
e
f
ir
s
t
f
u
lly
co
n
n
ec
ted
lay
e
r
s
will
r
ed
u
ce
th
e
d
im
en
s
io
n
ality
f
r
o
m
2
5
6
t
o
1
0
0
n
e
u
r
o
n
s
,
a
n
d
t
h
e
o
th
e
r
lay
er
will
d
eter
m
in
e
th
e
lik
elih
o
o
d
o
f
b
elo
n
g
in
g
to
a
v
ess
el
o
r
n
o
t [
1
2
]
.
Fig
u
r
e
1
.
Netwo
r
k
ar
ch
itectu
r
e
[
1
1
]
C
er
eb
r
o
v
ascu
lar
s
eg
m
en
tatio
n
m
eth
o
d
is
u
s
ed
to
an
aly
s
e
a
n
d
ev
alu
ate
s
o
m
e
T
OF
MRA
d
atasets
o
f
h
ea
lth
y
s
u
b
jects.
T
h
e
d
atasets
wer
e
ac
q
u
i
r
ed
f
o
r
t
h
e
s
eg
m
e
n
tatio
n
p
r
o
ce
s
s
.
Slab
b
o
u
n
d
a
r
y
ar
tef
ac
t
c
o
r
r
ec
tio
n
was
d
o
n
e
f
o
r
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
b
y
u
s
in
g
t
h
e
N3
alg
o
r
ith
m
[
1
3
]
,
in
ten
s
ity
n
o
n
-
u
n
if
o
r
m
ity
co
r
r
ec
tio
n
ca
n
also
b
e
d
o
n
e.
A
s
k
u
ll
s
tr
ip
p
in
g
alg
o
r
ith
m
[
1
4
]
is
also
u
s
ed
.
B
ased
o
n
th
e
p
r
e
-
p
r
o
ce
s
s
ed
T
OF
MRA
d
atasets
,
th
e
m
an
u
al
s
eg
m
en
tatio
n
o
f
th
e
v
e
s
s
els in
ea
ch
d
ataset
is
d
o
n
e.
T
h
e
p
atch
es f
r
o
m
all
d
ir
ec
tio
n
s
ar
e
ex
tr
ac
ted
f
r
o
m
a
cu
b
ic
r
eg
i
o
n
wh
ich
is
d
ef
in
ed
ar
o
u
n
d
all
v
o
x
els
in
s
id
e
th
e
b
r
ain
r
e
g
io
n
.
T
o
ca
lcu
late
th
e
v
ess
el
lik
elih
o
o
d
,
ea
ch
p
atch
is
f
ed
to
th
e
C
NN
[
1
5
]
.
T
h
e
n
,
f
o
r
ea
ch
o
r
ien
tatio
n
,
th
r
ee
p
r
o
b
a
b
ilit
y
m
ap
s
ar
e
p
r
esen
t.
Data
s
ets
co
n
s
is
tin
g
o
f
T
OF
MRA
im
ag
es
ar
e
r
an
d
o
m
ly
s
elec
ted
f
o
r
tr
ain
in
g
,
t
h
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
d
ee
p
C
NN
ca
n
b
e
ev
alu
ated
.
M
o
r
e
ac
c
u
r
ate
r
esu
lts
can
b
e
o
b
tain
e
d
if
mor
e
tr
ain
in
g
im
a
g
es
ar
e
u
s
ed
.
Fo
r
e
ac
h
im
ag
e
u
s
ed
f
o
r
test
in
g
,
th
e
tr
ain
in
g
im
ag
e
s
el
ec
ted
will
b
e
d
if
f
er
en
t.
T
h
en
,
t
h
er
e
is
n
ee
d
f
o
r
in
cr
ea
s
in
g
th
e
n
u
m
b
er
o
f
tr
ain
i
n
g
im
ag
es.
I
n
o
r
d
er
to
ev
alu
ate
t
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
ce
r
eb
r
o
v
ascu
lar
s
eg
m
en
tatio
n
u
s
in
g
C
NN
an
d
g
r
o
u
n
d
-
tr
u
th
m
an
u
al
s
eg
m
en
tatio
n
s
,
d
ice
s
im
ilar
ity
co
ef
f
icien
t
(
DSC
)
[
1
6
]
is
u
s
ed
.
DS
C
ca
n
b
e
ca
lcu
lated
as,
DSC
=
2
|
A
∩
B
|
/ (
|
A
|
+
|
B
|
)
,
wh
er
e
A
an
d
B
d
ef
in
es th
e
g
r
o
u
n
d
-
tr
u
t
h
an
d
C
NN
s
eg
m
en
tatio
n
s
,
r
esp
ec
tiv
ely
.
3
.
2
.
3
D
-
co
nv
o
lutio
na
l neura
l net
wo
r
k
(
3
D
-
CNN)
3D
-
C
NN
ar
ch
itectu
r
e
p
r
o
p
o
s
ed
b
y
Kan
d
il
et
a
l.
[
1
7
]
s
h
o
wn
in
Fig
u
r
e
2
,
c
o
n
s
is
ts
o
f
eig
h
t
co
n
v
o
l
u
tio
n
al
lay
er
s
,
two
f
u
ll
y
co
n
n
ec
ted
lay
e
r
s
,
an
d
o
n
e
cl
ass
if
icatio
n
lay
er
.
T
h
e
eig
h
t
lay
er
s
ar
e
h
av
in
g
3
0
,
3
0
,
4
0
,
4
0
,
4
0
,
4
0
,
5
0
,
5
0
f
ea
tu
r
e
m
ap
s
(
FMs)
an
d
t
h
e
k
er
n
e
l
s
ize
is
2
7
.
I
m
ag
e
s
eg
m
en
ts
with
s
ize
2
5
×2
5
×2
5
ar
e
u
s
ed
as in
p
u
t to
th
e
n
etwo
r
k
.
T
h
e
b
atch
s
ize
u
s
ed
is
1
0
s
e
g
m
en
ts
.
T
h
e
v
o
x
el
’
s
ex
ac
t p
o
s
itio
n
will b
e
lo
s
t if
p
o
o
lin
g
lay
er
is
p
r
esen
t
wh
ic
h
will
in
v
er
s
ely
a
f
f
ec
t
th
e
ac
c
u
r
ac
y
a
n
d
t
h
e
s
tr
id
es
ar
e
u
n
ar
y
.
T
h
e
PR
eL
u
n
o
n
-
lin
ea
r
ity
is
u
s
ed
b
y
t
h
is
3
D
C
NN
ar
ch
itectu
r
e
a
n
d
t
h
e
r
o
o
t
m
ea
n
s
q
u
ar
e
(
R
MS)
Pro
p
o
p
tim
izer
an
d
Nester
o
v
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
ma
tic
ce
r
eb
r
o
va
s
cu
la
r
s
eg
men
ta
tio
n
meth
o
d
s
-
a
r
ev
iew
(
F
a
tma
Ta
h
er
)
579
m
o
m
en
tu
m
with
v
al
u
es
L
1
=
1
0
−6
,
L
2
=
1
0
−4
a
n
d
m
=
0
.
6
ar
e
u
s
ed
f
o
r
tr
ain
in
g
.
T
h
e
le
ar
n
in
g
r
ate
an
d
th
e
d
r
o
p
o
u
t
ar
e
s
et
to
1
0
-
3
a
n
d
5
0
%
r
ate
r
esp
ec
tiv
ely
,
th
at
was
u
s
ed
o
n
th
e
las
t
h
id
d
en
lay
er
s
.
At
ea
ch
o
p
tim
izatio
n
s
tep
f
o
r
th
e
n
o
r
m
aliza
tio
n
o
f
t
h
e
FM
ac
tiv
atio
n
in
all
h
id
d
e
n
lay
er
s
,
b
atch
n
o
r
m
aliza
tio
n
tech
n
iq
u
e
was
u
s
ed
.
B
lo
o
d
f
l
o
w
s
ig
n
al
s
tr
en
g
t
h
in
s
id
e
t
h
e
b
r
ain
at
a
s
p
ec
if
ic
tim
e
v
ar
ie
s
f
r
o
m
o
n
e
ar
ea
t
o
an
o
th
er
.
I
n
o
r
d
er
to
f
ac
e
th
is
ch
allen
g
e,
in
s
id
e
ea
ch
c
o
m
p
a
r
tm
en
t,
b
l
o
o
d
v
ess
els
ar
e
o
f
s
am
e
f
ea
tu
r
es
wh
e
n
ea
ch
MRA v
o
lu
m
e
is
p
ar
titi
o
n
ed
in
to
two
c
o
m
p
ar
tm
e
n
ts
.
Fig
u
r
e
2
.
3
D
-
C
NN
ar
ch
itectu
r
e
Per
f
o
r
m
an
ce
o
f
th
e
s
eg
m
en
tat
io
n
p
r
o
ce
s
s
ca
n
b
e
en
h
a
n
ce
d
b
y
th
is
.
Du
r
in
g
th
e
p
a
r
titi
o
n
in
g
p
r
o
ce
s
s
,
ce
r
eb
r
al
b
io
-
m
a
r
k
er
ca
lled
ci
r
cle
o
f
W
illi
s
(
C
o
W
)
i
s
s
ele
cted
.
Mo
s
t
o
f
th
e
b
lo
o
d
v
ess
els
ar
e
o
f
d
if
f
er
en
t
d
iam
eter
s
ize
wh
en
it
is
e
x
is
ti
n
g
at
C
o
W
an
d
b
elo
w
it
an
d
it
is
h
av
in
g
s
m
all
d
iam
eter
s
ize
ab
o
v
e
C
o
W
.
B
ased
o
n
th
e
p
o
s
itio
n
o
f
th
e
MRA
s
lices,
wh
eth
er
it
is
ab
o
v
e
o
r
b
elo
w
C
o
W
ar
e
d
iv
id
ed
in
to
two
co
m
p
ar
tm
en
ts
.
All
th
e
b
lo
o
d
v
ess
els
s
h
o
u
ld
h
a
v
e
s
am
e
s
h
ap
e,
a
p
p
ea
r
a
n
ce
,
an
d
d
iam
eter
s
o
f
ap
p
r
o
x
i
m
ate
s
izes
in
ea
ch
co
m
p
ar
tm
en
t;
th
e
r
ef
o
r
e,
th
e
s
eg
m
en
tatio
n
ef
f
icien
cy
a
n
d
a
cc
u
r
ac
y
ca
n
b
e
in
cr
ea
s
ed
.
A
s
u
b
v
ascu
lar
tr
ee
is
p
r
o
d
u
ce
d
b
y
th
e
3
-
D
C
NN
m
an
ip
u
latio
n
d
u
r
in
g
t
h
is
p
r
o
ce
s
s
.
T
h
e
f
in
al
o
u
tc
o
m
e
is
o
b
tain
ed
b
y
co
m
b
in
in
g
tw
o
s
u
b
v
ascu
lar
t
r
ee
[
1
7
]
.
3
D
C
NN
s
eg
m
en
tatio
n
ac
c
u
r
ac
y
ca
n
b
e
test
ed
b
y
co
n
s
id
er
i
n
g
t
h
e
ev
alu
atio
n
m
etr
ic
DSC
.
I
n
th
e
ex
p
e
r
im
en
t
d
o
n
e
b
y
Kan
d
il
et
a
l.
[
1
7
]
,
tr
ain
in
g
s
et
co
n
s
is
ts
o
f
4
9
im
ag
es
a
n
d
test
in
g
is
d
o
n
e
f
o
r
1
7
im
ag
es.
3
.
3
.
3
D
U
-
Net
San
ch
es
[
1
]
p
r
o
p
o
s
ed
an
ar
c
h
it
ec
tu
r
e
ca
lled
Uce
p
tio
n
[
1
8
]
wh
ich
in
cr
ea
s
es
th
e
n
etwo
r
k
s
ize
b
y
ad
d
in
g
co
n
v
o
lu
tio
n
al
lay
er
s
o
r
b
y
in
cr
ea
s
in
g
t
h
eir
d
e
p
th
.
D
is
ad
v
an
tag
e
o
f
u
s
in
g
3
D
U
-
N
et
is
th
at
th
er
e
is
a
ch
an
ce
o
f
o
v
er
f
itti
n
g
wh
en
it
i
s
h
av
in
g
a
g
r
ea
ter
n
u
m
b
e
r
o
f
p
ar
am
eter
s
.
I
f
th
e
n
u
m
b
e
r
o
f
a
n
n
o
tated
im
a
g
es
is
less
in
b
io
m
ed
ical
d
atasets
,
th
is
p
r
o
b
lem
m
ai
n
ly
ar
is
es.
T
h
er
ef
o
r
e,
s
p
ar
s
ely
co
n
n
ec
te
d
ar
ch
itectu
r
es
ar
e
p
r
ef
er
r
e
d
.
I
n
t
h
is
s
p
ar
s
ely
co
n
n
ec
ted
ar
ch
itectu
r
e,
th
e
in
p
u
t
is
p
ass
ed
in
p
ar
allel
to
s
e
v
er
al
b
r
an
c
h
es
with
d
if
f
er
en
t
k
er
n
el
s
izes
an
d
f
i
n
a
lly
co
n
ca
ten
ated
in
t
h
e
en
d
.
I
n
ad
d
itio
n
,
th
is
ar
c
h
itectu
r
e
w
ith
1
D
co
n
v
o
lu
tio
n
will
d
is
s
o
ciate
th
e
d
ep
th
in
f
o
r
m
atio
n
in
th
e
ch
an
n
els
an
d
s
p
atial
in
f
o
r
m
atio
n
with
th
e
3
D
co
n
v
o
lu
tio
n
s
.
Mo
r
eo
v
er
,
p
r
o
ce
s
s
in
g
th
e
im
a
g
e
in
two
s
ca
les
is
p
o
s
s
ib
le
b
y
u
s
in
g
d
if
f
er
e
n
t
k
e
r
n
el
s
izes.
T
h
e
m
o
d
u
les
wh
ich
ar
e
cr
ea
ted
to
b
e
u
s
ed
in
a
3
D
m
o
d
el
ar
e:
th
e
f
ea
tu
r
es
m
ap
s
h
ap
e
is
r
etain
e
d
b
y
o
n
e
m
o
d
el
an
d
th
e
im
a
g
e
s
ize
is
m
ad
e
h
alf
with
p
ar
allel
s
tr
id
ed
co
n
v
o
l
u
tio
n
s
an
d
m
ax
p
o
o
lin
g
o
p
er
atio
n
s
b
y
t
h
e
o
th
er
m
o
d
el
to
m
a
k
e
co
n
tr
ac
tin
g
p
ath
[
1
]
.
I
n
th
e
e
x
p
an
s
iv
e
p
ath
,
u
p
-
s
am
p
lin
g
wa
s
u
s
ed
.
Uce
p
tio
n
ar
c
h
itectu
r
e
i
s
g
iv
en
in
Fig
u
r
e
3
.
Af
ter
ea
ch
co
n
v
o
lu
tio
n
lay
er
,
a
R
eL
u
ac
tiv
atio
n
f
u
n
ctio
n
i
s
ap
p
lied
.
A
s
ig
m
o
id
ac
tiv
ati
o
n
f
u
n
ctio
n
was
u
s
ed
af
ter
th
e
last
co
n
v
o
lu
tio
n
al
lay
e
r
.
T
h
e
r
ef
o
r
e
,
ea
ch
v
o
x
el
is
lin
k
e
d
to
a
p
r
o
b
ab
ilit
y
wh
ich
will
h
el
p
s
to
r
ea
ch
t
h
e
p
r
o
b
a
b
ilit
y
o
f
b
el
o
n
g
in
g
to
t
h
e
v
ess
els
n
etwo
r
k
d
u
r
in
g
th
e
tr
ain
in
g
.
Sin
ce
th
e
in
p
u
t
is
in
b
in
ar
y
s
eg
m
en
tatio
n
,
t
h
e
last
lay
e
r
h
a
s
o
n
ly
o
n
e
ch
a
n
n
el.
Af
ter
ea
c
h
ac
tiv
atio
n
f
u
n
ctio
n
,
d
r
o
p
o
u
t
was
u
s
ed
wh
en
th
e
r
eg
u
lar
izatio
n
tec
h
n
iq
u
e
g
o
es.
T
h
e
n
eg
ativ
e
o
f
th
e
d
ice
co
e
f
f
i
cien
t is u
s
ed
f
o
r
th
e
lo
s
s
f
u
n
cti
o
n
,
f
r
o
m
(
1
)
:
(
,
)
=
2
∗
|
∩
|
|
|
+
|
|
(
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
10
,
N
o
.
3
,
Sep
tem
b
er
202
1
:
5
7
6
-
583
580
w
h
er
e
P
is
th
e
p
r
ed
ictio
n
o
f
th
e
n
etwo
r
k
a
n
d
T
is
th
e
g
r
o
u
n
d
tr
u
th
.
C
o
n
v
er
g
en
c
e
s
p
ee
d
is
h
ig
h
an
d
is
n
u
m
er
ically
s
tab
le
as
its
v
alu
e
is
b
etwe
en
ze
r
o
an
d
o
n
e.
B
y
u
s
in
g
th
e
b
ac
k
-
p
r
o
p
ag
atio
n
alg
o
r
ith
m
with
th
e
Ad
am
o
p
tim
izer
,
th
e
tr
ain
in
g
was
d
o
n
e.
Mo
r
eo
v
er
,
cu
b
ic
o
f
p
atch
es
o
f
6
4
×
64
×
6
4
v
o
x
els
wer
e
u
s
ed
to
f
ee
d
th
e
d
ata.
Fo
r
v
alid
atio
n
,
n
o
n
-
s
u
p
er
p
o
s
ed
p
atch
es
wer
e
u
s
ed
.
I
n
ad
d
itio
n
,
m
o
r
e
d
ata
is
o
b
t
ain
ed
as
a
r
esu
lt
o
f
tr
ain
in
g
th
r
o
u
g
h
p
atch
es,
s
in
ce
d
ata
au
g
m
e
n
tatio
n
was
n
o
t
u
s
ed
;
o
n
ly
3
6
im
ag
es
f
r
o
m
th
is
d
ataset
wer
e
d
is
p
o
s
ed
.
A
tech
n
iq
u
e
ca
lled
s
n
ap
s
h
o
t
en
s
em
b
le
[
1
9
]
is
u
s
e
d
to
en
h
an
ce
th
e
g
en
er
aliza
tio
n
b
y
a
v
er
ag
in
g
th
e
weig
h
ts
o
f
th
e
s
am
e
m
o
d
el
at
d
if
f
er
e
n
t
m
o
m
e
n
ts
o
f
t
h
e
tr
ai
n
in
g
.
Du
r
in
g
tr
ain
in
g
with
a
cy
clic
lear
n
in
g
r
ate
s
ch
ed
u
le,
th
ese
m
o
m
e
n
ts
ar
e
c
h
o
s
en
as th
e
lo
ca
l m
in
im
a
o
f
t
h
e
v
alid
atio
n
l
o
s
s
.
Fig
u
r
e
3
.
Uce
p
tio
n
ar
c
h
itectu
r
e
[
1
]
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
W
h
en
C
NN
ap
p
r
o
ac
h
[
2
0
]
is
u
s
ed
f
o
r
s
eg
m
en
tatio
n
o
f
b
lo
o
d
v
ess
els
in
t
im
e
o
f
f
lig
h
t
MR
A
im
ag
es,
DSC
v
alu
es o
f
f
iv
e
d
ataset
s
ar
e
av
er
ag
ed
an
d
o
b
tain
ed
0
.
7
6
4
±
0
.
0
1
0
.
Fro
m
Ph
ellan
et
a
l.
[
1
2
]
ex
p
er
im
e
n
t,
it is
p
r
o
v
e
d
th
at
as
a
g
r
ea
ter
n
u
m
b
er
o
f
im
a
g
es
u
s
ed
will
in
cr
ea
s
e
th
e
tr
ain
in
g
tim
es
an
d
t
h
e
ti
m
e
tak
in
g
f
o
r
test
in
g
ar
e
n
o
t
d
ep
e
n
d
in
g
o
n
th
e
q
u
an
tity
o
f
tr
ain
in
g
im
ag
es
an
d
th
e
s
eg
m
en
tatio
n
ac
cu
r
ac
y
a
n
d
c
an
n
o
t
b
e
in
cr
ea
s
ed
b
y
m
o
r
e
t
r
ain
in
g
im
ag
es.
Fig
u
r
e
4
s
h
o
ws
th
e
s
eg
m
en
tatio
n
r
esu
lts
u
s
in
g
th
e
d
ee
p
C
N
N
[
2
1
]
.
Fig
u
r
e
4
(
a)
s
h
o
ws
th
e
r
esu
lt
o
b
tain
ed
m
a
n
u
ally
an
d
Fig
u
r
e
4
(
b
)
d
ep
ict
s
th
e
C
NN
ap
p
r
o
ac
h
o
u
tp
u
t.
Fo
r
3
D
-
C
NN
[
2
2
]
,
s
eg
m
en
tatio
n
ac
cu
r
a
cy
ca
n
b
e
d
eter
m
in
ed
u
s
in
g
a
co
m
m
o
n
s
eg
m
en
t
atio
n
ev
alu
atio
n
m
etr
ic
ca
lled
d
ice
s
im
ilar
ity
co
ef
f
icien
t (
DSC
)
is
u
s
ed
.
(
a)
(
b
)
Fig
u
r
e
4
.
Seg
m
e
n
tatio
n
r
esu
lts
u
s
in
g
C
NN
:
(
a)
m
an
u
al
a
n
d
(
b
)
o
u
t
p
u
t o
f
C
NN
ap
p
r
o
ac
h
T
h
is
s
eg
m
en
tatio
n
m
eth
o
d
s
h
o
ws
g
o
o
d
p
er
f
o
r
m
an
ce
in
c
h
ar
ac
ter
izin
g
ce
r
e
b
r
al
v
ascu
l
atu
r
e
with
0
.
8
3
2
±
2
.
3
0
DSC
wh
en
c
o
m
p
ar
ed
with
m
a
n
u
ally
s
eg
m
e
n
te
d
g
r
o
u
n
d
tr
u
t
h
.
E
ac
h
MRA
v
o
lu
m
e
is
p
ar
titi
o
n
e
d
ab
o
v
e
C
o
W
an
d
b
elo
w
C
o
W
to
p
er
f
o
r
m
th
e
s
eg
m
en
tatio
n
[
2
3
]
task
lo
ca
lly
o
v
er
th
e
en
tir
e
b
r
ain
.
T
h
e
r
esu
lts
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
ma
tic
ce
r
eb
r
o
va
s
cu
la
r
s
eg
men
ta
tio
n
meth
o
d
s
-
a
r
ev
iew
(
F
a
tma
Ta
h
er
)
581
o
f
s
eg
m
en
tatio
n
s
h
o
u
ld
b
e
p
o
s
itiv
e,
as th
e
o
th
er
ev
alu
atio
n
p
ar
am
eter
s
o
f
th
e
e
x
tr
ac
ted
f
ea
t
u
r
es a
r
e
in
f
lu
e
n
ce
d
[
2
4
]
.
3
D
-
C
NN
s
eg
m
en
tatio
n
r
esu
lt
is
s
h
o
wn
in
Fig
u
r
e
5
.
Fig
u
r
e
5
(
a
)
s
h
o
ws
th
e
r
esu
lt
o
b
t
ain
ed
m
an
u
ally
a
n
d
Fig
u
r
e
5
(
b
)
d
e
p
icts
th
e
3
D
-
C
NN
ap
p
r
o
ac
h
o
u
tp
u
t.
(
a)
(
b
)
Fig
u
r
e
5
.
Seg
m
e
n
tatio
n
r
esu
lts
u
s
in
g
3
D
-
C
NN
:
(
a)
g
r
o
u
n
d
tr
u
th
,
(
b
)
o
u
t
p
u
t o
f
3
D
-
C
NN
ap
p
r
o
ac
h
B
y
u
s
in
g
3
D
U
-
Net
[
2
5
]
a
p
p
r
o
ac
h
,
DSC
o
b
tain
e
d
wh
e
n
co
m
p
ar
ed
with
th
e
g
r
o
u
n
d
tr
u
th
is
0
.
6
7
1
±
0
.
0
1
.
Fig
u
r
e
6
s
h
o
ws
th
e
s
eg
m
e
n
tatio
n
r
esu
lts
o
f
3
D
U
-
Net
wh
en
co
m
p
ar
ed
with
its
g
r
o
u
n
d
-
tr
u
th
.
Fig
u
r
e
6
(
a)
s
h
o
ws th
e
r
esu
lt o
b
tain
ed
m
an
u
ally
an
d
Fig
u
r
e
6
(
b
)
d
e
p
ic
ts
th
e
3
D
U
-
Net
ap
p
r
o
ac
h
o
u
tp
u
t.
Fin
ally
,
th
ese
m
eth
o
d
s
ar
e
c
o
m
p
ar
ed
with
th
e
g
lo
b
al
s
tatis
t
ical
-
b
ased
ap
p
r
o
ac
h
(
GSB
)
p
r
o
p
o
s
ed
b
y
El
-
B
az
et
a
l
.
[
2
6
]
a
n
d
th
e
DSC
v
alu
e
o
b
tain
e
d
in
t
h
is
m
eth
o
d
is
0
.
8
0
1
±
2
.
7
.
Fig
u
r
e
7
s
h
o
ws
th
e
s
eg
m
en
tatio
n
r
esu
lt
u
s
in
g
GSB
.
Fig
u
r
e
7
(
a)
s
h
o
ws
th
e
r
esu
lt
o
b
tain
ed
m
an
u
ally
an
d
Fig
u
r
e
7
(
b
)
d
e
p
icts
th
e
GSB
ap
p
r
o
ac
h
o
u
tp
u
t.
(
a)
(
b
)
Fig
u
r
e
6
.
Seg
m
e
n
tatio
n
r
esu
lts
u
s
in
g
3
D
U
-
Net
:
(
a)
g
r
o
u
n
d
tr
u
th
an
d
(
b
)
o
u
tp
u
t o
f
Uce
p
tio
n
ap
p
r
o
ac
h
(
a)
(
b
)
Fig
u
r
e
7
.
Seg
m
e
n
tatio
n
r
esu
lts
u
s
in
g
GSB
:
(
a)
g
r
o
u
n
d
tr
u
th
a
n
d
(
b
)
o
u
t
p
u
t o
f
GSB
ap
p
r
o
ac
h
C
o
m
p
ar
in
g
th
e
DC
S
v
alu
es,
we
ar
e
ab
le
to
f
in
d
th
at
th
e
p
er
f
o
r
m
an
ce
o
f
3
D
U
-
Net
is
h
ig
h
er
co
m
p
ar
ed
to
o
th
e
r
m
eth
o
d
s
.
3
D
U
-
Net
s
h
o
ws
less
s
im
i
lar
i
ty
with
th
e
m
an
u
al
m
eth
o
d
w
h
ich
p
r
o
v
es
it
as
a
b
etter
m
eth
o
d
.
T
h
e
DSC
v
alu
es
o
b
tain
ed
f
o
r
th
e
t
h
r
ee
s
eg
m
en
tatio
n
m
et
h
o
d
s
wh
en
c
o
m
p
ar
in
g
with
th
e
g
r
o
u
n
d
tr
u
t
h
ar
e
s
h
o
w
n
in
T
a
b
le
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
10
,
N
o
.
3
,
Sep
tem
b
er
202
1
:
5
7
6
-
583
582
T
ab
le
1
.
DSC
v
alu
es b
ased
o
n
th
e
ty
p
e
o
f
s
eg
m
en
tatio
n
u
s
ed
Ty
p
e
s
o
f
s
e
g
m
e
n
t
a
t
i
o
n
D
S
C
v
a
l
u
e
C
N
N
0
.
7
6
4
±
0
.
0
1
3D
-
C
N
N
0
.
8
3
2
±
2
.
3
0
3
D
U
-
N
e
t
G
S
B
0
.
6
7
1
±
0
.
0
1
0
.
8
0
1
±
2
.
7
5.
CO
NCLU
SI
O
N
No
wad
ay
s
,
ce
r
eb
r
o
v
ascu
lar
d
i
s
ea
s
es
ar
e
b
ec
o
m
in
g
o
n
e
o
f
th
e
m
ajo
r
ca
u
s
es
f
o
r
i
n
cr
ea
s
in
g
th
e
d
ea
th
r
ate
o
f
th
e
wo
r
ld
p
o
p
u
latio
n
.
Diag
n
o
s
is
an
d
test
in
g
o
f
s
u
ch
d
is
ea
s
es
ar
e
s
till
a
ch
alle
n
g
e
f
o
r
h
ea
lth
ca
r
e
p
r
o
v
id
e
r
s
.
T
h
e
r
ef
o
r
e
,
s
eg
m
en
t
atio
n
o
f
th
e
ce
r
eb
r
o
v
asc
u
lar
s
t
r
u
ctu
r
e
is
a
v
er
y
im
p
o
r
tan
t
s
tep
f
o
r
th
e
d
iag
n
o
s
es
p
r
o
ce
s
s
.
C
o
m
p
u
ter
-
aid
ed
d
ia
g
n
o
s
is
(
C
AD)
s
y
s
tem
i
s
u
s
ed
b
y
th
e
clin
ician
s
in
th
e
p
r
ed
ictio
n
p
r
o
ce
s
s
.
I
n
th
is
p
ap
er
,
s
o
m
e
o
f
s
u
ch
au
to
m
ati
c
s
eg
m
en
tatio
n
m
eth
o
d
s
lik
e
C
NN,
3
D
-
C
NN
an
d
3
D
U
-
Ne
t
ar
e
ex
p
la
in
ed
a
n
d
co
m
p
ar
ed
th
eir
p
e
r
f
o
r
m
an
ce
.
I
n
th
is
p
ap
er
,
s
o
m
e
o
f
s
u
ch
au
to
m
atic
s
eg
m
en
tatio
n
m
eth
o
d
s
lik
e
C
N
N,
3
D
-
C
NN
an
d
3
D
U
-
Net
ar
e
ex
p
lain
ed
an
d
co
m
p
ar
ed
th
eir
p
er
f
o
r
m
an
ce
.
I
n
th
e
C
NN
ap
p
r
o
ac
h
,
o
n
ly
a
s
m
aller
n
u
m
b
er
o
f
s
eg
m
en
te
d
g
r
o
u
n
d
tr
u
th
im
ag
es
ar
e
n
ee
d
ed
to
o
b
tain
g
o
o
d
r
esu
lts
wh
ich
m
ak
es
th
is
ap
p
licatio
n
ea
s
ier
in
m
a
n
y
p
lace
s
s
u
ch
as
in
r
esear
ch
o
r
clin
ical
f
ield
.
3
D
C
NN
is
v
er
y
h
elp
f
u
l
i
n
t
h
e
s
eg
m
en
tatio
n
o
f
s
m
all
as
well
as
co
m
p
lex
b
l
o
o
d
v
ess
el.
Fo
r
eith
er
h
ea
lth
y
o
r
u
n
h
ea
lth
y
v
ess
els,
th
is
m
e
th
o
d
is
a
p
p
licab
le.
Ho
wev
er
,
C
NN
s
h
o
ws
b
etter
p
er
f
o
r
m
an
ce
t
h
an
3
D
-
C
NN.
3
D
U
-
Net
ar
ch
itectu
r
e
is
p
r
o
p
o
s
ed
as
a
r
esu
lt
o
f
th
e
in
s
p
ir
atio
n
f
r
o
m
th
e
in
ce
p
tio
n
ar
ch
itectu
r
e.
W
h
en
c
o
m
p
ar
ed
with
th
e
o
r
ig
in
al
U
-
Net,
t
h
is
m
o
d
el
s
h
o
wed
b
etter
p
er
f
o
r
m
an
ce
.
L
o
s
s
f
u
n
c
tio
n
n
ee
d
s
to
b
e
s
elec
ted
with
m
u
ch
ca
r
e;
th
er
ef
o
r
e,
th
e
to
ta
l
n
u
m
b
er
o
f
v
o
x
els
in
th
e
d
ata
is
n
o
t
c
o
n
s
id
er
ed
b
y
th
e
d
ice
co
ef
f
icien
t.
B
y
co
m
p
ar
in
g
t
h
e
th
r
ee
s
eg
m
en
tatio
n
m
eth
o
d
s
,
3
D
U
-
Net
s
h
o
wed
th
e
b
etter
p
er
f
o
r
m
an
ce
.
RE
F
E
R
E
NC
E
S
[1
]
P
.
S
a
n
c
h
e
sa
,
C.
M
e
y
e
r,
V.
Vi
g
o
n
,
a
n
d
B
.
Na
e
g
e
l,
"
Ce
re
b
ro
v
a
sc
u
lar
Ne
two
rk
S
e
g
m
e
n
tatio
n
o
f
M
RA
Im
a
g
e
s
with
De
e
p
Lea
rn
in
g
,
"
2
0
1
9
IEE
E
1
6
t
h
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
siu
m
o
n
Bi
o
me
d
ica
l
Im
a
g
i
n
g
(I
S
BI
2
0
1
9
)
,
Ve
n
ice
,
Italy
,
2
0
1
9
,
p
p
.
7
6
8
-
7
7
1
,
d
o
i:
1
0
.
1
1
0
9
/I
S
BI.
2
0
1
9
.
8
7
5
9
5
6
9
.
[2
]
D.
S
a
lo
n
e
r,
“
Th
e
AA
P
M
/RS
NA
p
h
y
sic
s
t
u
t
o
rial
f
o
r
re
sid
e
n
ts
-
a
n
i
n
tro
d
u
c
ti
o
n
to
M
R
a
n
g
i
o
g
ra
p
h
y
,
”
Ra
d
io
g
ra
p
h
ics
,
v
o
l.
1
5
,
n
o
.
2
,
p
p
.
4
5
3
-
4
6
5
,
1
9
9
5
,
DO
I:
d
o
i
.
o
r
g
/1
0
.
1
1
4
8
/rad
i
o
g
ra
p
h
i
c
s.1
5
.
2
.
7
7
6
1
6
4
8
.
[3
]
S
.
F
a
n
,
Y.
Bian
,
H.
Ch
e
n
,
a
n
d
Y
.
Ka
n
g
,
“
Un
su
p
e
rv
ise
d
c
e
re
b
r
o
v
a
s
c
u
lar
se
g
m
e
n
tatio
n
o
f
TOF
-
M
RA
ima
g
e
s
b
a
se
d
o
n
d
e
e
p
n
e
u
ra
l
n
e
two
rk
a
n
d
h
id
d
e
n
m
a
rk
o
v
ra
n
d
o
m
field
m
o
d
e
l,
”
Fro
n
ti
e
rs
in
n
e
u
ro
i
n
f
o
rm
a
ti
c
s
,
2
0
2
0
,
p
p
.
1
-
1
0
,
d
o
i:
d
o
i.
o
r
g
/
1
0
.
3
3
8
9
/fn
i
n
f.
2
0
1
9
.
0
0
0
7
7
.
[4
]
C.
M
ish
ra
a
n
d
D.L
.
G
u
p
ta,
“
De
e
p
M
a
c
h
in
e
Lea
rn
in
g
a
n
d
Ne
u
ra
l
Ne
two
rk
s:
A
n
Ov
e
rv
iew
”
,
IAE
S
In
t
e
rn
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
i
a
l
I
n
telli
g
e
n
c
e
(
IJ
AI)
,
v
o
l.
6
,
n
o
.
2
,
p
p
.
6
6
-
7
3
,
J
u
n
e
2
0
1
7
.
[5
]
C.
Ch
e
n
,
C.
Qi
n
,
H.
Q
u
i,
a
n
d
G
.
Tarro
n
i,
“
De
e
p
lea
rn
i
n
g
f
o
r
c
a
rd
i
a
c
ima
g
e
se
g
m
e
n
tatio
n
-
a
re
v
iew
,
”
a
rXiv
2
0
1
9
,
p
p
.
1
-
4
7
,
d
o
i:
1
0
.
3
3
8
9
/fcv
m
.
2
0
2
0
.
0
0
0
2
5
.
[6
]
M
.
He
sa
m
ian
a
n
d
W
.
Jia
,
“
De
e
p
lea
rn
in
g
tec
h
n
i
q
u
e
s
f
o
r
m
e
d
ica
l
ima
g
e
se
g
m
e
n
tatio
n
,
Ac
h
iev
e
m
e
n
ts
a
n
d
c
h
a
ll
e
n
g
e
s,”
J
o
u
r
n
a
l
o
f
d
i
g
it
a
l
im
a
g
i
n
g
,
p
p
.
5
8
2
-
5
9
6
,
2
0
1
9
,
d
o
i:
1
0
.
1
0
0
7
/s
1
0
2
7
8
-
0
1
9
-
0
0
2
2
7
-
x.
[7
]
O.
Cice
k
,
A.
Ab
d
u
lk
h
a
d
ir,
a
n
d
S
.
S
.
Li
e
n
k
a
m
p
,
“
3
D
U
-
Ne
t:
Lea
rn
in
g
d
e
n
se
v
o
lu
m
e
tri
c
se
g
m
e
n
tatio
n
fr
o
m
sp
a
rse
a
n
n
o
tati
o
n
,
”
a
rXiv
,
2
0
1
6
,
p
p
.
1
-
8
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
4
6
7
2
3
-
8
_
4
9
.
[8
]
M
.
Li
v
n
e
,
J.
Re
i
g
e
r,
O.
U.
Ay
d
i
n
,
a
n
d
A.
A.
Tah
a
,
“
A
U
-
Ne
t
d
e
e
p
lea
rn
in
g
fra
m
e
wo
rk
o
f
h
i
g
h
p
e
rf
o
rm
a
n
c
e
v
e
ss
e
l
se
g
m
e
n
tatio
n
i
n
p
a
ti
e
n
ts
with
c
e
re
b
ro
v
a
sc
u
lar
d
ise
a
se
,
”
Fro
n
ti
e
rs
in
n
e
u
ro
sc
ien
c
e
,
v
o
l.
1
3
,
p
p
.
1
-
1
3
,
2
0
1
9
,
d
o
i
:
1
0
.
3
3
8
9
/f
n
i
n
s.2
0
1
9
.
0
0
0
9
7
.
[9
]
X.
G
a
o
a
n
d
Y.
Uc
h
i
y
a
m
a
,
“
A
fa
st
a
n
d
fu
l
ly
a
u
t
o
m
a
ti
c
m
e
th
o
d
f
o
r
c
e
re
b
ro
v
a
sc
u
lar
se
g
m
e
n
tat
io
n
o
n
ti
m
e
-
of
-
fl
ig
h
t
M
RA i
m
a
g
e
,
’’
J
o
u
r
n
a
l
o
f
d
i
g
it
a
l
ima
g
i
n
g
,
v
o
l.
2
4
,
p
p
.
6
0
9
-
6
2
5
,
Au
g
u
st
2
0
1
1
,
d
o
i:
1
0
.
1
0
0
7
/s1
0
2
7
8
-
0
1
0
-
9
3
2
6
-
1.
[1
0
]
F
.
Tah
e
r,
A.
M
a
h
m
o
u
d
,
A.
S
h
a
lab
y
,
a
n
d
A.
El
-
Ba
z
,
“
A
Re
v
iew
o
n
th
e
Ce
re
b
r
o
v
a
sc
u
lar
S
e
g
m
e
n
tatio
n
M
e
th
o
d
s”
,
2
0
1
8
I
EE
E
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
s
iu
m
o
n
S
i
g
n
a
l
Pr
o
c
e
ss
in
g
a
n
d
I
n
fo
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
(I
S
S
PI
T
)
,
Lo
u
is
v
il
le,
KY
,
USA,
2
0
1
8
,
p
p
.
3
5
9
-
3
6
4
,
d
o
i:
1
0
.
1
1
0
9
/IS
S
P
IT.
2
0
1
8
.
8
6
4
2
7
5
6
.
[1
1
]
F
.
Tah
e
r
e
t
a
l
.
,
“
Ac
c
u
r
a
te
S
e
g
m
e
n
tatio
n
o
f
Ce
re
b
r
o
v
a
sc
u
latu
r
e
fro
m
TOF
-
M
RA
Im
a
g
e
s
Us
i
n
g
A
p
p
e
a
ra
n
c
e
De
sc
rip
to
rs,” in
I
EE
E
Acc
e
ss
,
v
o
l
.
8
,
p
p
.
9
6
1
3
9
-
9
6
1
4
9
,
2
0
2
0
,
d
o
i:
1
0
.
1
1
0
9
/ACCES
S
.
2
0
2
0
.
2
9
8
2
8
6
9
.
[1
2
]
R.
P
h
e
ll
a
n
,
A.
P
e
ix
i
n
h
o
,
A.
F
a
lca
o
,
a
n
d
N.
F
o
r
k
e
rt,
“
Va
sc
u
lar se
g
m
e
n
tatio
n
o
f
TOF
M
RA
ima
g
e
s o
f
th
e
b
ra
in
u
sin
g
a
d
e
e
p
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
,
”
S
p
ri
n
g
e
r
,
p
p
.
3
9
-
4
6
,
2
0
1
7
,
d
o
i:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
6
7
5
3
4
-
3
_
5
.
[1
3
]
J.
G
.
S
led
,
A.
P
.
Zi
j
d
e
n
b
o
s,
a
n
d
A.
C
.
E
v
a
n
s,
“
A
n
o
n
p
a
ra
m
e
tri
c
m
e
th
o
d
fo
r
a
u
t
o
m
a
ti
c
c
o
rre
c
ti
o
n
o
f
in
ten
si
t
y
n
o
n
u
n
ifo
rm
it
y
i
n
M
RI
d
a
ta”
,
in
I
EE
E
T
ra
n
sa
c
ti
o
n
s
o
n
M
e
d
ic
a
l
Im
a
g
i
n
g
,
v
o
l.
1
7
,
n
o
.
1
,
p
p
.
8
7
-
9
7
,
F
e
b
.
1
9
9
8
,
d
o
i:
1
0
.
1
1
0
9
/
4
2
.
6
6
8
6
9
8
.
[1
4
]
N.
F
o
r
k
e
rt,
D
.
S
a
ri
n
g
,
a
n
d
J.
F
.
I
ll
ies
,
“
Au
to
m
a
ti
c
b
ra
in
se
g
m
e
n
tat
io
n
in
ti
m
e
-
of
-
fli
g
h
t
M
RA
ima
g
e
s,”
M
e
th
o
d
s
o
f
in
fo
rm
a
ti
o
n
i
n
me
d
icin
e
,
v
o
l.
4
8
,
n
o
.
5
,
p
p
.
3
9
9
-
4
0
7
,
2
0
0
9
.
[1
5
]
K.
Ka
m
n
it
sa
s,
L.
Ch
e
n
,
a
n
d
C.
Led
ig
,
“
M
u
l
ti
-
sc
a
le
3
D
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
f
o
r
les
io
n
s
e
g
m
e
n
tatio
n
in
b
ra
in
M
RI
,
”
S
e
ma
n
ti
c
sc
h
o
la
r
,
2
0
1
5
,
p
p
.
1
-
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
u
to
ma
tic
ce
r
eb
r
o
va
s
cu
la
r
s
eg
men
ta
tio
n
meth
o
d
s
-
a
r
ev
iew
(
F
a
tma
Ta
h
er
)
583
[1
6
]
L.
R.
Dic
e
,
“
M
e
a
su
re
s
o
f
t
h
e
a
m
o
u
n
t
o
f
e
c
o
lo
g
ic
a
ss
o
c
iatio
n
b
e
t
we
e
n
sp
e
c
ies
,
”
Eco
lo
g
y
,
p
p
.
2
9
7
-
3
0
2
,
1
9
4
5
,
d
o
i:
d
o
i.
o
rg
/1
0
.
2
3
0
7
/1
9
3
2
4
0
9
.
[1
7
]
H.
Ka
n
d
il
e
t
a
l
.
,
“
A
n
o
v
e
l
c
o
m
p
u
ter
-
a
id
e
d
d
ia
g
n
o
sis
sy
ste
m
fo
r
th
e
e
a
rly
d
e
tec
ti
o
n
o
f
h
y
p
e
rte
n
sio
n
b
a
se
d
o
n
c
e
re
b
ro
v
a
sc
u
lar
a
lt
e
ra
ti
o
n
s,”
Ne
u
ro
Ima
g
e
:
Cli
n
ica
l,
El
se
v
ier
,
p
.
1
0
2
1
0
7
,
v
o
l
.
2
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j.
n
icl
.
2
0
1
9
.
1
0
2
1
0
7
.
[1
8
]
C.
S
z
e
g
e
d
y
e
t
a
l
.
,
“
G
o
in
g
d
e
e
p
e
r
with
c
o
n
v
o
lu
t
io
n
s”
,
2
0
1
5
IEE
E
Co
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
Vi
si
o
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
io
n
(CVP
R)
,
Bo
st
o
n
,
M
A,
USA,
2
0
1
5
,
p
p
.
1
-
9
,
d
o
i
:
1
0
.
1
1
0
9
/CVP
R.
2
0
1
5
.
7
2
9
8
5
9
4
.
[1
9
]
G
.
Hu
a
n
g
,
Y.
Li
,
G
.
P
leiss
,
Z.
Li
u
,
a
n
d
J.
E
.
Ho
p
c
ro
ft,
“
S
n
a
p
s
h
o
t
e
n
se
m
b
les
:
Trai
n
1
,
g
e
t
M
f
o
r
fre
e
,
”
Co
RR
,
a
b
s/1
7
0
4
.
0
0
1
0
9
,
2
0
1
7
.
[2
0
]
F
.
S
u
lt
a
n
a
,
A.
S
u
fian
,
P
.
Du
tt
a
,
“
Ev
o
l
u
ti
o
n
o
f
Im
a
g
e
S
e
g
m
e
n
tati
o
n
u
sin
g
De
e
p
Co
n
v
o
lu
ti
o
n
a
l
Ne
tw
o
rk
:
A
S
u
r
v
e
y
”
,
a
rXiv
,
p
p
.
1
-
3
8
,
M
a
y
2
0
2
0
.
[2
1
]
R.
P
o
o
jary
,
R.
Ra
i
n
a
,
A.
K.
M
o
n
d
a
l,
“
Eff
e
c
t
o
f
d
a
ta
-
a
u
g
m
e
n
tatio
n
o
n
fin
e
-
t
u
n
e
d
CNN
m
o
d
e
l
p
e
rf
o
rm
a
n
c
e
”
,
IAE
S
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
Art
if
icia
l
In
telli
g
e
n
c
e
,
v
o
l.
1
0
,
n
o
.
1
,
p
p
.
8
4
-
9
2
,
M
a
rc
h
2
0
2
1
.
[2
2
]
Y.
He
,
H.
Yu
,
X.
Li
u
,
Z.
Ya
n
g
,
W.
S
u
n
,
Y.
Wan
g
,
Q.
F
u
,
Y.
Z
o
u
,
“
De
e
p
Lea
rn
in
g
b
a
se
d
3
D
S
e
g
m
e
n
tatio
n
:
A
S
u
rv
e
y
”
,
a
rXiv
,
p
p
.
1
-
3
6
,
M
a
rc
h
2
0
2
1
.
[2
3
]
A.
El
-
Ba
z
,
A.
fa
ra
g
,
a
n
d
G
.
G
im
e
lfarb
,
“
Ce
re
b
ro
v
a
sc
u
lar
se
g
m
e
n
t
a
ti
o
n
b
y
a
c
c
u
ra
te
p
r
o
b
a
b
i
li
stic
m
o
d
e
li
n
g
o
f
TOF
-
M
RA i
m
a
g
e
s,”
S
p
ri
n
g
e
r
,
2
0
0
5
,
p
p
.
1
1
2
8
-
1
1
3
7
,
d
o
i:
1
0
.
1
0
0
7
/
1
1
4
9
9
1
4
5
_
1
.
[2
4
]
A.
Aja
m
,
A.
A.
Az
iz,
V.
S
.
As
irv
a
d
a
m
,
A.
S
.
M
u
d
a
,
I.
F
a
y
e
,
a
n
d
S
.
J.
S
a
fd
a
r
G
a
rd
e
z
i,
“
A
Re
v
iew
o
n
S
e
g
m
e
n
tatio
n
a
n
d
M
o
d
e
li
n
g
o
f
Ce
re
b
ra
l
Va
sc
u
l
a
tu
re
fo
r
S
u
r
g
ica
l
P
lan
n
in
g
,
”
in
I
EE
E
Acc
e
ss
,
v
o
l.
5
,
p
p
.
1
5
2
2
2
-
1
5
2
4
0
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
0
9
/ACCE
S
S
.
2
0
1
7
.
2
7
1
8
5
9
0
.
[2
5
]
O.
Ro
n
n
e
b
e
rg
e
r,
P
.
F
isc
h
e
r,
T.
Bro
x
,
“
U
-
Ne
t
-
Co
n
v
o
lu
ti
o
n
a
l
N
e
two
rk
s
fo
r
Bi
o
m
e
d
ica
l
Im
a
g
e
S
e
g
m
e
n
tatio
n
”
,
a
rXiv
,
p
p
.
1
-
8
,
M
a
y
2
0
1
5
.
[2
6
]
A.
El
-
Ba
z
e
t
a
l
.
,
“
P
re
c
ise
S
e
g
m
e
n
tatio
n
o
f
3
-
D
M
a
g
n
e
ti
c
Re
so
n
a
n
c
e
An
g
io
g
ra
p
h
y
”
,
in
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Bi
o
me
d
ica
l
En
g
in
e
e
rin
g
,
v
o
l.
5
9
,
n
o
.
7
,
p
p
.
2
0
1
9
-
2
0
2
9
,
J
u
ly
2
0
1
2
,
d
o
i:
1
0
.
1
1
0
9
/T
BM
E.
2
0
1
2
.
2
1
9
6
4
3
4
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
Fa
tm
a
Ta
h
e
r
is
a
n
a
ss
o
c
iate
p
r
o
fe
ss
o
r
a
n
d
a
ss
istan
t
d
e
a
n
f
o
r
re
se
a
rc
h
a
n
d
o
u
trea
c
h
in
th
e
Co
ll
e
g
e
o
f
Tec
h
n
o
l
o
g
ica
l
I
n
n
o
v
a
t
io
n
a
t
Zay
e
d
Un
i
v
e
rsity
,
D
u
b
a
i,
UA
E.
He
r
re
se
a
rc
h
in
tere
sts
a
re
in
t
h
e
a
re
a
s
o
f
si
g
n
a
l
a
n
d
ima
g
e
p
ro
c
e
ss
in
g
,
p
a
tt
e
rn
re
c
o
g
n
it
i
o
n
,
d
e
e
p
lea
rn
in
g
,
M
a
c
h
in
e
lea
rn
in
g
,
a
rti
ficia
l
in
tell
ig
e
n
c
e
,
m
e
d
ica
l
ima
g
e
a
n
a
ly
sis,
e
sp
e
c
ially
i
n
d
e
tec
ti
n
g
o
f
t
h
e
c
a
n
c
e
ro
u
s
c
e
ll
s,
k
id
n
e
y
tra
n
sp
lan
t
a
n
d
a
u
ti
sm
.
In
a
d
d
it
io
n
to
t
h
a
t,
wa
term
a
rk
in
g
,
re
m
o
te
se
n
sin
g
,
a
n
d
sa
telli
te
ima
g
e
s
re
se
a
rc
h
e
s.
Dr
.
F
a
tma
Tah
e
r
h
a
s
p
u
b
li
s
h
e
d
m
o
re
th
a
n
7
0
p
a
p
e
rs
in
in
tern
a
ti
o
n
a
l
Jo
u
rn
a
ls
a
n
d
c
o
n
fe
re
n
c
e
s.
S
h
e
se
rv
e
d
a
s
a
m
e
m
b
e
r
o
f
th
e
ste
e
rin
g
,
o
rg
a
n
izin
g
a
n
d
tec
h
n
ica
l
p
ro
g
ra
m
c
o
m
m
it
tee
s
o
f
m
a
n
y
in
ter
n
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s,
a
n
d
h
a
s
se
rv
e
d
o
n
m
a
n
y
e
d
it
o
rial
a
n
d
re
v
iew
in
g
b
o
a
rd
s
o
f
in
tern
a
ti
o
n
a
l
j
o
u
r
n
a
ls
a
n
d
c
o
n
fe
re
n
c
e
s.
Dr.
F
a
tma
is
th
e
c
h
a
ir
o
f
IEE
E
UA
E
se
c
ti
o
n
a
n
d
t
h
e
c
h
a
ir
o
f
th
e
E
d
u
c
a
ti
o
n
C
o
m
m
it
tee
in
Brit
ish
S
o
c
iety
in
UA
E.
Dr.
F
a
tma
h
a
s
re
c
e
iv
e
d
m
a
n
y
d
isti
n
g
u
ish
e
d
a
wa
rd
s
s
u
c
h
a
s:
Be
st
p
a
p
e
r
a
wa
rd
o
f
t
h
e
first
p
rize
in
th
e
P
h
D
F
o
r
u
m
o
f
th
e
2
0
th
I
E
EE
In
ter
n
a
ti
o
n
a
l
Co
n
fe
re
n
c
e
o
n
El
e
c
tro
n
ics
,
Circu
it
s,
a
n
d
S
y
ste
m
s
(ICECS
),
P
h
D
F
o
ru
m
,
De
c
.
8
-
1
1
,
2
0
1
3
.
An
d
re
c
e
n
tl
y
,
sh
e
wa
s
a
wa
rd
e
d
th
e
UA
E
P
io
n
e
e
rs
a
wa
rd
a
s
t
h
e
first
UA
E
to
c
re
a
te
a
c
o
m
p
u
ter
a
id
e
d
d
iag
n
o
sis
sy
ste
m
f
o
r
e
a
rly
lu
n
g
c
a
n
c
e
r
d
e
tec
ti
o
n
b
a
se
d
o
n
th
e
sp
u
tu
m
c
o
lo
r
ima
g
e
a
n
a
ly
sis,
a
wa
rd
e
d
b
y
H.H
S
h
e
ik
M
o
h
a
m
m
e
d
Bin
Ra
sh
e
d
Al
M
a
k
to
u
m
,
1
5
t
h
N
o
v
.
2
0
1
5
.
In
a
d
d
i
ti
o
n
to
th
a
t
s
h
e
wa
s
a
wa
rd
e
d
a
n
in
n
o
v
a
ti
o
n
a
wa
rd
a
t
th
e
2
0
1
6
Emira
ti
W
o
m
e
n
Aw
a
rd
s
b
y
H.
H.
S
h
e
ik
Ah
m
e
d
Bin
S
a
e
e
d
Al
-
M
a
k
t
o
u
m
.
Ch
a
irma
n
o
f
Ci
v
il
A
v
iatio
n
Au
t
h
o
rit
y
,
a
n
d
Lo
re
a
l
-
UN
ES
CO
fo
r
Wo
m
e
n
in
S
c
ien
c
e
M
i
d
d
l
e
Eas
t
F
e
ll
o
ws
h
i
p
2
0
1
7
.
Ne
e
m
a
Pra
k
a
sh
is
c
u
rre
n
tl
y
w
o
rk
in
g
a
s
re
se
a
rc
h
a
ss
istan
t
in
t
h
e
Co
ll
e
g
e
o
f
Tec
h
n
o
l
o
g
ica
l
In
n
o
v
a
ti
o
n
a
t
Zay
e
d
Un
i
v
e
rsity
,
Du
b
a
i,
UA
E.
S
h
e
re
c
e
iv
e
d
M
.
T
ech
.
d
e
g
re
e
in
Op
t
o
e
lec
tro
n
ic
s
a
n
d
Op
ti
c
a
l
c
o
m
m
u
n
ica
ti
o
n
fr
o
m
Un
iv
e
rsity
o
f
Ke
ra
la,
In
d
ia
(2
0
1
6
),
B.
tec
h
d
e
g
re
e
i
n
El
e
c
tro
n
ics
a
n
d
C
o
m
m
u
n
ica
ti
o
n
fro
m
M
a
h
a
tma
G
a
n
d
h
i
Un
i
v
e
rsity
,
Ke
ra
la,
I
n
d
ia
(
2
0
1
2
)
.
He
r
re
se
a
rc
h
in
tere
sts
a
re
in
th
e
a
re
a
s
o
f
ima
g
e
p
ro
c
e
ss
in
g
,
d
e
sig
n
in
g
e
lec
tro
n
ic
c
ircu
i
ts
a
n
d
tele
c
o
m
m
u
n
ica
ti
o
n
.
S
h
e
h
a
s
g
a
in
e
d
e
x
p
e
rien
c
e
in
ima
g
e
p
r
o
c
e
ss
in
g
wo
rk
in
g
in
Ra
m
a
n
Re
se
a
rc
h
In
stit
u
te,
In
d
ia
a
n
d
h
a
s
g
o
o
d
e
x
p
e
rien
c
e
in
tes
ti
n
g
e
lec
tro
n
ic
c
ircu
it
s
a
n
d
c
o
m
p
o
n
e
n
ts,
wo
r
k
i
n
g
i
n
I
n
d
ia
n
S
p
a
c
e
Re
se
a
rc
h
Org
a
n
iza
t
io
n
,
I
n
d
i
a
.
Evaluation Warning : The document was created with Spire.PDF for Python.