I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
10
,
No
.
6
,
Dec
em
b
er
202
0
,
p
p
.
6
6
8
2
~6
6
9
0
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
0
i
6
.
p
p
6
6
8
2
-
6
6
9
0
6682
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m/in
d
ex
.
p
h
p
/I
JE
C
E
Funct
io
na
l
m
a
g
n
etic
r
eso
na
nce i
ma
g
ing
-
ba
sed bra
in decodin
g
w
ith
v
isua
l se
m
a
n
tic
m
o
del
P
iy
a
wa
t
Sa
eng
pet
ch,
L
uepo
l
P
ipa
n
m
e
m
e
k
a
po
rn
,
S
uw
a
t
c
ha
i K
a
m
o
ls
a
ntir
o
j
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter an
d
I
n
f
o
rm
a
ti
o
n
S
c
ien
c
e
,
Kin
g
M
o
n
g
k
u
t'
s Un
iv
e
rsit
y
o
f
Tec
h
n
o
lo
g
y
No
rth
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
7
,
2
0
1
9
R
ev
i
s
ed
Ma
y
31
,
2020
A
cc
ep
ted
J
u
n
15
,
2
0
2
0
T
h
e
a
c
ti
v
it
y
p
a
tt
e
rn
o
f
th
e
b
ra
in
h
a
s
b
e
e
n
a
c
ti
v
a
ted
to
id
e
n
ti
f
y
a
p
e
rso
n
i
n
m
in
d
.
Us
in
g
th
e
f
u
n
c
ti
o
n
m
a
g
n
e
ti
c
re
so
n
a
n
c
e
im
a
g
in
g
(f
M
RI
)
to
d
e
c
ip
h
e
r
b
ra
in
d
e
c
o
d
in
g
is
th
e
m
o
st
a
c
c
e
p
ted
m
e
th
o
d
.
Ho
we
v
e
r,
th
e
a
c
c
u
ra
c
y
o
f
f
M
RI
-
b
a
se
d
b
ra
in
d
e
c
o
d
e
r
is
stil
l
re
stricte
d
d
u
e
to
li
m
it
e
d
train
in
g
sa
m
p
les
.
T
h
e
li
m
it
a
ti
o
n
s
o
f
th
e
b
ra
in
d
e
c
o
d
e
r
u
sin
g
f
M
RI
a
r
e
p
a
ss
e
d
th
ro
u
g
h
th
e
d
e
sig
n
f
e
a
tu
re
s
p
ro
p
o
se
d
f
o
r
m
a
n
y
lab
e
l
c
o
d
in
g
a
n
d
m
o
d
e
l
t
ra
in
in
g
to
p
re
d
ict
th
e
se
c
h
a
ra
c
teristics
f
o
r
a
p
a
rti
c
u
lar
lab
e
l.
M
o
re
o
v
e
r,
w
h
a
t
k
in
d
o
f
se
m
a
n
ti
c
fe
a
tu
re
s
f
o
r
d
e
c
ip
h
e
rin
g
th
e
n
e
u
ro
l
o
g
ica
l
a
c
ti
v
it
y
p
a
tt
e
rn
s
a
re
u
n
c
lea
r.
In
c
u
rre
n
t
w
o
rk
,
a
n
e
w
c
a
lcu
latio
n
m
o
d
e
l
f
o
r
lea
rn
in
g
d
e
c
o
d
in
g
lab
e
ls
th
a
t
is
c
o
n
siste
n
t
w
it
h
f
M
RI
a
c
ti
v
it
y
r
e
sp
o
n
se
s
.
T
h
e
a
p
p
ro
a
c
h
d
e
m
o
n
stra
tes
th
e
p
ro
p
o
se
d
c
o
r
re
sp
o
n
d
in
g
lab
e
l'
s
su
c
c
e
ss
in
term
s
o
f
a
c
c
u
ra
c
y
,
w
h
ich
is
d
e
c
o
d
e
d
f
ro
m
b
ra
in
a
c
ti
v
it
y
p
a
tt
e
rn
s
a
n
d
c
o
m
p
a
re
d
w
it
h
c
o
n
v
e
n
ti
o
n
a
l
tex
t
-
d
e
riv
e
d
f
e
a
tu
r
e
tec
h
n
iq
u
e
.
Be
sid
e
s,
e
x
p
e
rime
n
tal
stu
d
ies
p
re
se
n
t
a
train
in
g
m
o
d
e
l
b
a
se
d
o
n
m
u
lt
i
-
tas
k
in
g
to
re
d
u
c
e
th
e
p
ro
b
lem
s
o
f
li
m
it
e
d
train
in
g
d
a
ta
se
ts.
T
h
e
re
fo
re
,
th
e
m
u
lt
i
-
tas
k
lea
rn
in
g
m
o
d
e
l
is
m
o
re
e
ff
ici
e
n
t
th
a
n
m
o
d
e
rn
m
e
th
o
d
s
o
f
c
a
lcu
latio
n
,
a
n
d
d
e
c
o
d
i
n
g
f
e
a
tu
re
s
m
a
y
b
e
e
a
sil
y
o
b
tain
e
d
.
K
ey
w
o
r
d
s
:
B
r
ain
d
ec
o
d
in
g
Dee
p
lear
n
in
g
f
MRI
ac
ti
v
it
y
p
atter
n
s
Mu
lti
-
ta
s
k
lear
n
in
g
Vis
u
a
l se
m
an
t
ic
m
o
d
el
Co
p
y
rig
h
t
©
2
0
2
0
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
P
iy
a
w
at
Saen
g
p
etch
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
an
d
I
n
f
o
r
m
a
tio
n
Scie
n
ce
,
Kin
g
Mo
n
g
k
u
t
's Un
i
v
er
s
i
t
y
o
f
T
ec
h
n
o
lo
g
y
,
No
r
th
B
an
g
k
o
k
,
B
an
g
k
o
k
,
T
h
ailan
d
.
E
m
ail
:
s
5
7
0
4
0
6
2
9
1
0
0
1
7
@
k
m
u
tn
b
.
ac
.
t
h
1.
I
NT
RO
D
UCT
I
O
N
R
esear
ch
ab
o
u
t
u
n
d
er
s
ta
n
d
in
g
b
r
ain
f
u
n
ctio
n
s
b
ased
o
n
th
e
r
elatio
n
s
h
i
p
b
et
w
ee
n
th
o
u
g
h
t
an
d
s
ti
m
u
la
tio
n
o
f
th
e
b
r
ain
.
T
h
e
p
o
s
s
ib
ilit
y
to
u
n
d
er
s
tan
d
th
e
b
r
ain
'
s
f
u
n
ctio
n
p
atter
n
is
s
ti
m
u
l
ated
b
y
th
i
n
g
s
s
u
c
h
as
s
o
u
n
d
,
s
m
ell,
li
g
h
t,
o
r
o
b
j
e
cts.
No
w
t
h
er
e
ar
e
m
ed
ical
a
n
al
y
s
i
s
to
o
ls
t
h
at
ca
n
g
et
d
ata
f
r
o
m
th
e
b
r
ai
n
s
u
ch
as
co
m
p
u
t
ed
to
m
o
g
r
ap
h
y
(
C
T
)
,
elec
tr
o
en
ce
p
h
alo
g
r
ap
h
y
(
E
E
G)
an
d
m
a
g
n
eto
e
n
ce
p
h
alo
g
r
ap
h
y
(
ME
G)
to
m
o
tio
n
b
r
ai
n
f
u
n
ctio
n
s
.
F
u
n
ct
io
n
m
a
g
n
etic
r
eso
n
an
ce
i
m
ag
i
n
g
(
f
M
R
I
)
is
a
tech
n
iq
u
e
th
at
allo
w
s
t
h
e
b
r
ain
to
d
is
tin
g
u
is
h
b
et
w
ee
n
t
h
e
s
tate
s
o
f
m
i
n
d
.
T
h
e
f
MRI
i
m
a
g
es
ar
e
g
e
n
e
r
ated
f
r
o
m
b
lo
o
d
-
o
x
y
g
e
n
-
lev
el
s
d
ep
en
d
e
n
t
(
B
OL
D)
s
ti
m
u
latio
n
o
f
b
r
ain
ce
lls
.
Mo
s
t
o
f
t
h
e
r
esear
ch
h
as
b
ee
n
d
o
n
e
o
n
b
r
ain
tr
a
n
s
cr
ip
tio
n
w
i
th
f
M
R
I
,
s
u
c
h
as
b
r
ain
-
co
m
p
u
ter
-
i
n
ter
f
ac
e
(
BC
I
)
,
n
eu
r
al
-
co
n
tr
o
l
i
n
t
er
f
ac
e
(
NC
I
)
is
a
d
ir
ec
t
li
n
e
o
f
co
m
m
u
n
icatio
n
b
e
t
w
e
e
n
t
h
e
b
r
a
i
n
g
e
t
s
w
i
r
e
d
o
r
i
m
p
r
o
v
e
m
e
n
t
a
n
d
e
x
t
e
r
n
a
l
d
e
v
i
c
e
s
[
1
]
,
a
s
l
i
e
d
e
t
e
c
t
i
o
n
,
e
x
p
o
s
i
n
g
t
h
e
d
e
l
i
b
e
r
a
t
e
d
ec
ep
tio
n
f
r
o
m
t
h
e
te
x
t
o
f
t
h
e
s
p
ee
ch
.
Fa
ls
e
d
etec
t
io
n
b
y
lie
d
etec
to
r
i
s
f
r
a
u
d
d
etec
tio
n
b
y
ev
a
lu
at
in
g
th
e
co
n
ten
t
o
f
m
es
s
ag
e
s
o
r
in
clu
d
i
n
g
id
en
ti
f
ier
s
[
2
]
.
B
r
ain
in
j
u
r
y
o
r
tr
au
m
atic
b
r
ain
i
n
j
u
r
y
th
at
ca
n
b
e
m
ea
s
u
r
ed
ex
ter
n
a
ll
y
[
3
]
.
T
h
e
3
-
d
im
e
n
s
io
n
al
i
m
a
g
e
o
b
tain
ed
f
r
o
m
f
M
R
I
co
n
s
is
ts
o
f
a
b
r
ain
p
o
in
t
ca
lled
v
o
x
el
(
v
o
lu
m
e
+
p
ix
el
)
.
A
v
o
x
el
s
h
o
w
s
b
r
ai
n
ac
tiv
it
y
,
w
h
ich
i
n
d
icate
t
h
e
ar
ea
s
o
f
th
e
b
r
ai
n
ar
e
w
o
r
k
i
n
g
.
E
ac
h
v
o
x
el
s
h
o
w
s
a
p
ar
t
o
f
th
e
b
r
ain
,
co
n
tain
in
g
b
r
ain
ce
ll
s
,
o
n
e
m
illi
o
n
ce
l
ls
in
t
h
e
m
in
d
o
f
t
h
e
s
u
b
j
ec
t.
T
h
e
b
r
ain
ac
ti
v
it
y
i
s
ca
p
tu
r
ed
b
y
id
e
n
ti
f
y
in
g
d
i
f
f
e
r
en
t
ac
ti
v
ed
v
o
x
el
s
.
I
n
g
e
n
e
r
al,
th
e
f
o
r
m
at
o
f
d
ec
o
d
in
g
b
r
ain
u
s
in
g
f
MRI
e
m
p
h
a
s
ized
th
e
lin
ea
r
r
elatio
n
s
h
ip
b
et
w
ee
n
t
h
e
s
tate
an
d
th
e
m
i
n
d
th
r
o
u
g
h
s
ta
tis
tica
l
test
s
.
I
n
th
e
liter
atu
r
e
[
4
]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
F
u
n
ctio
n
ma
g
n
etic
r
eso
n
a
n
ce
ima
g
in
g
-
b
a
s
ed
b
r
a
in
d
ec
o
d
i
n
g
w
ith
visu
a
l sema
n
tic
…
(
P
iya
w
a
t
S
a
en
g
p
etc
h
)
6683
p
r
o
p
o
s
ed
th
e
g
en
er
al
l
in
ea
r
m
o
d
el
(
GL
M)
ap
p
r
o
ac
h
,
h
o
wev
er
,
th
i
s
e
x
a
m
in
in
g
ac
ti
v
ed
v
o
x
el
s
ap
p
r
o
ac
h
is
is
o
latio
n
.
Mo
v
eo
v
er
a
m
u
lt
i
-
v
o
x
el
p
atter
n
s
o
f
ac
ti
v
it
y
(
MV
P
A
)
[
5
]
is
s
tr
o
n
g
tech
n
iq
u
e
a
n
d
w
a
s
p
r
o
p
o
s
ed
t
o
d
etec
t
f
M
R
I
v
o
x
el
s
.
S
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM)
a
n
d
lin
ea
r
d
is
cr
i
m
i
n
a
n
t
a
n
al
y
s
i
s
(
L
D
A
)
a
lg
o
r
it
h
m
is
g
en
er
all
y
u
s
ed
f
o
r
tr
ain
i
n
g
m
o
d
el
s
,
a
s
u
b
s
et
o
f
t
h
e
f
MR
I
d
ata
d
ec
o
d
in
g
.
A
lth
o
u
g
h
t
h
e
m
ac
h
in
e
lear
n
i
n
g
alg
o
r
ith
m
h
as
alr
ea
d
y
ac
h
iev
ed
th
e
ef
f
icie
n
c
y
o
f
MV
P
A
,
its
ac
cu
r
ac
y
is
lo
w
d
u
e
to
t
h
e
lack
o
f
tr
ain
i
n
g
ex
a
m
p
le
s
.
T
h
e
is
s
u
es
w
il
l
b
e
c
o
n
f
i
n
ed
to
th
e
r
ep
r
esen
tatio
n
o
f
class
e
s
ap
p
lied
f
o
r
tr
ain
in
g
,
w
h
ile
t
y
p
e
s
th
at
d
o
n
o
t
b
e
in
tr
ain
i
n
g
ca
n
n
o
t
b
e
d
ec
o
d
ed
.
Fro
m
th
e
li
m
ita
tio
n
s
o
f
in
f
o
r
m
atio
n
f
o
r
th
e
tr
ai
n
ed
m
o
d
el,
th
er
e
f
o
r
e
s
ea
r
ch
i
n
g
f
o
r
an
d
s
u
g
g
esti
n
g
n
e
w
m
e
th
o
d
s
.
I
n
d
esi
g
n
in
g
f
e
atu
r
es
f
o
r
d
ec
o
d
in
g
th
e
b
r
ain
t
h
at
d
o
es
n
o
t
ex
i
s
t
i
n
th
e
li
m
ited
s
et.
P
io
n
ee
r
liter
atu
r
e
[
6
]
s
tates
t
h
at
b
r
ain
f
u
n
ctio
n
p
atter
n
s
ca
n
b
e
p
r
ed
ic
ted
u
s
i
n
g
s
e
m
an
ti
c
r
elatio
n
s
h
ip
s
l
in
k
i
n
g
2
5
-
v
er
b
s
an
d
n
a
m
es.
P
alatu
cc
i
e
t
al.
[
7
]
h
a
v
e
i
n
v
esti
g
ated
a
2
1
8
-
d
i
m
e
n
s
io
n
a
l
r
ep
r
esen
tatio
n
o
f
t
h
e
s
u
b
j
ec
ts
r
ec
eiv
ed
f
r
o
m
2
1
8
v
o
lu
n
t
ee
r
s
w
h
o
w
er
e
a
n
s
w
er
ed
t
h
e
q
u
esti
o
n
s
r
elate
d
to
th
e
ca
teg
o
r
y
o
f
o
b
j
ec
ts
s
u
ch
as
"
I
s
it
c
o
ld
?
"
o
r
"
I
s
it
h
o
t?
"
o
r
"
C
a
n
yo
u
w
a
lk
o
n
it?
"
.
So
m
e
s
t
u
d
ies
h
a
v
e
s
u
g
g
e
s
ted
a
m
ec
h
an
is
m
f
o
r
t
h
e
au
to
m
atic
e
x
tr
ac
tio
n
o
f
m
ea
n
in
g
b
y
e
x
a
m
in
in
g
la
n
g
u
a
g
e
k
n
o
w
led
g
e
s
u
ch
a
s
W
ik
ip
ed
ia
[
8
]
an
d
W
o
r
d
Net
[
9
]
.
E
x
iti
n
g
r
esear
ch
u
s
ed
th
e
o
u
ts
ta
n
d
in
g
f
ea
t
u
r
es
f
r
o
m
t
h
e
te
x
t
to
d
ec
ip
h
er
th
e
b
r
ain
,
w
h
ic
h
is
f
o
u
n
d
to
b
e
a
p
er
f
ec
t
m
et
h
o
d
.
I
n
th
is
r
e
s
ea
r
ch
,
w
e
p
r
esen
t
th
e
v
is
u
al
f
ea
t
u
r
es
ass
o
ciate
d
w
it
h
t
h
e
class
o
f
o
b
j
ec
ts
,
w
h
ic
h
in
cr
ea
s
e
s
ac
cu
r
ac
y
.
T
h
is
m
e
th
o
d
ch
o
o
s
es
th
e
i
m
a
g
es
r
elat
ed
to
th
e
o
b
j
ec
ts
in
th
e
s
a
m
e
clas
s
a
n
d
d
if
f
er
e
n
t
c
h
ar
ac
ter
is
tic
s
,
s
u
c
h
as
i
m
a
g
e
s
ize,
b
r
ig
h
tn
e
s
s
,
o
r
ie
n
tatio
n
.
T
h
er
ef
o
r
e,
w
e
co
n
s
id
er
m
ed
i
u
m
an
d
h
i
g
h
-
le
v
el
i
m
ag
e
p
r
o
p
er
ti
es
i
n
e
n
co
d
in
g
m
an
y
o
b
j
ec
ts
.
W
e
m
ea
s
u
r
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
v
is
u
al
f
e
atu
r
es
b
y
u
s
in
g
co
lo
r
h
is
to
g
r
a
m
s
a
n
d
co
r
r
elo
g
r
a
m
s
.
T
h
e
o
n
li
n
e
i
m
ag
e
l
ib
r
ar
y
I
m
ag
eNe
t
h
a
s
i
n
cl
u
d
ed
i
m
a
g
e
s
an
d
ta
g
g
ed
i
m
ag
e
s
.
W
e
t
h
en
to
o
k
p
ictu
r
es
to
ex
tr
ac
t
t
h
e
f
e
atu
r
es
f
o
r
a
m
o
d
el
in
p
r
ed
ictin
g
ac
tiv
it
y
i
n
t
h
e
b
r
ain
o
f
b
r
ain
ac
tiv
it
y
d
ata
r
ec
o
r
d
ed
f
r
o
m
f
M
R
I
s
tu
d
ie
s
b
as
ed
o
n
m
u
l
ti
-
tas
k
i
n
g
lear
n
in
g
,
i
n
cl
u
d
in
g
i
n
-
d
ep
th
lear
n
in
g
to
cr
ea
te
f
ea
t
u
r
es
a
n
d
tr
an
s
f
er
lear
n
i
n
g
o
f
d
ee
p
lear
n
i
n
g
s
u
ch
a
s
R
esNet5
0
,
VG
G1
6
.
W
e
u
s
e
1
5
0
im
a
g
e
s
f
r
o
m
I
m
a
g
eNe
t
t
o
cr
ea
te
f
ea
t
u
r
es.
W
e
th
e
n
d
e
s
ig
n
ed
a
m
o
d
el
f
o
r
b
r
ain
p
r
ed
ictio
n
w
it
h
v
i
s
u
a
l
s
e
m
a
n
tic
f
ea
tu
r
e
s
an
d
f
M
R
I
d
ata
w
ith
m
u
lti
-
tas
k
i
n
g
lear
n
in
g
to
p
r
ed
ict
6
0
co
n
ce
p
ts
.
T
h
e
r
esu
lts
o
f
t
h
is
r
esear
ch
s
h
o
w
th
e
m
o
d
el
'
s
e
f
f
i
cien
c
y
a
n
d
p
r
ed
ictiv
e
ca
p
ab
ili
t
y
w
h
en
co
m
p
ar
ed
w
it
h
t
h
e
s
tate
-
of
-
t
h
e
-
ar
t.
T
h
e
m
ai
n
co
n
tr
ib
u
t
io
n
s
ar
e
as
f
o
llo
w
s
:
T
o
d
ec
o
d
e
th
e
b
r
ain
u
s
in
g
i
m
a
g
es
f
r
o
m
f
M
R
I
.
W
e
ex
p
lo
r
e
th
e
o
u
ts
tan
d
i
n
g
f
ea
t
u
r
es
f
r
o
m
i
m
ag
e
s
r
elate
d
to
th
e
o
b
j
ec
ts
o
r
im
ag
e
s
to
i
m
p
r
o
v
e
th
e
ac
c
u
r
ac
y
a
n
d
co
m
p
ar
ed
w
it
h
t
h
e
p
r
ev
io
u
s
r
e
s
ea
r
ch
.
T
o
p
r
esen
t
th
e
m
o
d
els
th
a
t
ar
e
u
s
ed
to
d
ec
ip
h
er
ac
t
iv
i
ties
i
n
t
h
e
b
r
ain
b
y
u
s
i
n
g
m
u
lt
ip
le
task
r
eg
r
es
s
io
n
m
o
d
el
s
an
d
d
e
m
o
n
s
tr
ati
n
g
th
e
s
u
itab
il
it
y
o
f
p
r
ed
ictin
g
p
r
o
p
er
ties
o
f
o
b
j
ec
ts
w
it
h
o
u
t t
h
e
tr
ai
n
in
g
s
e
t.
T
h
e
r
em
ain
d
e
r
o
f
th
is
d
o
cu
m
en
t
h
as
th
e
f
o
l
lo
w
in
g
s
tr
u
ctu
r
e
.
T
h
e
s
e
c
o
n
d
s
ec
ti
o
n
r
ev
iew
s
th
e
p
r
o
b
lem
o
f
b
r
ain
t
r
an
s
c
r
i
p
ti
o
n
,
al
o
n
g
w
ith
th
e
m
o
d
el
o
f
b
r
ain
t
r
an
s
cr
i
p
t
io
n
.
T
h
e
th
i
r
d
s
ec
t
io
n
p
r
es
en
ts
th
e
im
ag
e
f
ea
tu
r
es.
T
h
e
p
r
o
p
o
s
ed
m
o
d
e
l
is
d
es
c
r
i
b
e
d
in
Sec
ti
o
n
4
.
Secti
o
n
5
p
r
es
en
t
th
e
ex
p
e
r
i
m
en
tal
s
etu
p
,
an
d
th
e
r
esu
lts
a
ch
iev
e
d
an
d
d
is
cu
s
s
io
n
in
Se
cti
o
n
6
.
T
h
e
s
u
m
m
ar
y
an
d
c
o
n
clu
s
i
o
n
o
f
th
e
p
r
es
en
tati
o
n
in
s
e
cti
o
n
7
.
2.
B
RAIN
T
RA
NSCR
I
P
T
I
O
N
2
.
1
.
Understa
nd
ing
bra
in de
co
din
g
I
n
th
i
s
p
ath
,
w
e
d
escr
ib
e
th
e
b
r
ain
d
ec
o
d
in
g
p
r
o
b
lem
b
y
a
p
p
l
y
i
n
g
3
D
i
m
a
g
es
o
b
tai
n
ed
f
r
o
m
th
e
f
M
R
I
d
ataset.
T
h
e
i
m
ag
e
f
o
r
f
MRI
d
ata
X
=
×
m
atr
i
x
its
(
i,
j
)
th
elem
e
n
t
,
co
m
p
r
is
in
g
d
i
m
en
s
io
n
al
o
f
f
M
R
I
d
ata
in
ea
ch
r
o
w
.
T
h
e
f
u
n
ctio
n
ca
teg
o
r
y
to
p
r
ed
ict
a
lab
el.
I
n
g
e
n
er
al,
th
e
class
if
ic
a
tio
n
m
o
d
el
ca
n
b
e
ex
p
r
ess
ed
as
(
)
=
+
,
w
h
er
e
is
th
e
ca
lcu
late
d
er
r
o
r
.
T
h
en
is
w
ei
g
h
t
d
ir
ec
tio
n
.
T
h
e
d
if
f
ic
u
lt
y
is
to
f
i
n
d
to
d
ec
r
ea
s
e
=
−
(
)
an
d
th
en
i
s
f
o
r
m
u
late
t
h
e
w
a
y
o
u
t
t
o
th
e
o
p
ti
m
izat
io
n
i
n
tr
icac
y
ca
n
b
e
d
ef
in
ed
as
:
∗
=
1
2
‖
−
‖
2
2
+
(
)
(
1
)
w
h
er
e
(
)
s
y
m
b
o
lizes
t
h
e
p
ar
t
ex
p
er
ien
ce
o
f
th
e
c
h
alle
n
g
e
a
n
d
r
eg
u
lar
izes
t
h
e
m
o
d
el
.
I
n
th
e
t
y
p
ical
ca
s
e,
g
iv
e
n
≤
,
w
h
er
e
p
o
s
s
ib
le
o
b
j
e
cts
ar
e
e
x
clu
d
ed
f
r
o
m
t
h
e
tr
a
in
i
n
g
s
et,
co
n
s
tr
ai
n
i
n
g
th
e
o
u
tp
u
ts
f
o
r
d
ec
r
y
p
ti
n
g
.
B
ec
au
s
e
it
is
h
ar
d
to
g
et
a
n
f
MRI
i
m
a
g
e
f
o
r
ev
er
y
p
o
s
s
ib
le
o
b
j
ec
t,
p
ast
r
esear
ch
ca
n
o
n
l
y
d
ec
ip
h
er
th
e
i
n
f
o
r
m
atio
n
tau
g
h
t.
T
h
er
ef
o
r
e,
th
i
s
s
o
l
u
tio
n
s
o
lv
e
s
t
h
e
a
b
o
v
e
p
r
o
b
lem
s
f
o
r
d
ata
th
at
h
as
n
e
v
er
b
ee
n
s
ee
n
b
ef
o
r
e
an
d
s
u
p
p
o
r
ts
d
ata
ex
p
an
s
io
n
.
2
.
2
.
Repre
s
ent
a
t
io
n
m
o
dels
o
f
bra
in deco
din
g
B
r
ain
d
ec
o
d
in
g
is
p
r
esen
ted
u
tili
zin
g
s
em
an
tic
f
ea
tu
r
es
to
o
b
tain
th
e
lab
el
s
et
f
o
r
n
ew
in
p
u
t
d
ata.
R
eg
u
lar
d
ata
s
ets
o
b
tain
ed
f
r
o
m
ex
p
er
ien
ce
,
m
ea
n
in
g
f
o
r
d
esc
r
ib
in
g
th
e
m
o
s
t
lik
ely
co
n
ce
p
ts
Fig
u
r
e
1
.
B
r
ain
Dec
o
d
in
g
Mo
d
el.
I
t
is
co
m
p
o
s
ed
o
f
3
lev
els
:
1
)
Vo
x
els
ac
tiv
atio
n
v
ec
to
r
=
{
1
,
2
,
…
,
}
is
r
ep
r
esen
ted
in
th
e
in
p
u
t
lay
er
is
o
b
tain
ed
f
MRI
p
r
o
ce
s
s
in
g
;
2
)
Sem
an
tic
f
ea
tu
r
e
v
ec
to
r
is
r
ep
r
esen
ted
in
in
ter
m
ed
iate
lay
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Dec
em
b
er
2
0
2
0
:
6
6
8
2
-
6
6
9
0
6684
f
o
r
d
escr
ib
in
g
o
b
j
ec
ts
=
{
1
,
2
,
…
,
}
;
an
d
3
)
P
o
s
s
ib
le
o
b
j
ec
ts
=
{
1
,
2
,
…
,
}
is
r
ep
r
esen
ted
in
th
e
d
ec
o
d
in
g
lay
er
.
T
h
e
co
r
r
elatio
n
b
etw
ee
n
an
d
is
lear
n
ed
b
y
u
s
in
g
an
f
MRI
s
u
b
s
et
f
o
r
p
r
ed
ictin
g
th
e
f
ea
tu
r
e
v
alu
es,
w
h
er
ea
s
th
e
f
o
r
ec
ast f
ea
tu
r
es
ar
e
u
s
ed
f
o
r
d
ec
o
d
in
g
th
e
co
r
r
elate
d
o
b
j
ec
t
[
1
0
]
.
Fig
u
r
e
1
.
B
r
ain
d
ec
o
d
in
g
m
o
d
el
3.
I
M
AG
E
F
E
A
T
UR
E
S
C
r
ea
tu
r
es
ten
d
to
th
in
k
o
u
t,
"
I
s
it
co
ld
?
"
o
r
"
I
s
it
h
o
t?
"
,
o
r
"
C
a
n
yo
u
w
a
lk
o
n
it?
"
.
an
d
"
Ho
w
d
o
yo
u
feel
w
h
en
viewin
g
th
e
ima
g
e?
"
.
I
t
is
th
u
s
th
e
ca
p
ab
ilit
y
to
id
en
tif
y
th
e
p
o
s
s
ib
le
item
s
b
ased
o
n
s
tan
d
ar
d
v
is
u
al
f
ea
tu
r
es.
I
n
th
is
p
ar
t,
w
e
w
an
t
to
co
n
s
tr
u
ct
s
em
an
tic
f
ea
tu
r
es
f
r
o
m
im
ag
es
to
d
escr
ib
e
ac
tiv
ities
th
at
o
cc
u
r
in
th
e
b
r
ain
,
b
y
f
in
d
in
g
v
ar
io
u
s
m
eth
o
d
s
in
d
escr
ib
in
g
th
e
im
ag
es
r
elate
d
to
th
e
co
n
ce
p
t.
I
t
w
as
th
en
cr
ea
tin
g
a
m
o
d
el
o
f
m
ed
icin
e,
th
e
ty
p
e
o
f
ac
tiv
ity
in
th
e
b
r
ain
.
3
.
1
.
H
ier
a
rc
hica
l
v
is
ua
l f
ea
t
ures
I
m
p
le
m
e
n
ti
n
g
h
ier
ar
ch
ical
lea
r
n
in
g
in
w
h
ic
h
th
i
s
m
et
h
o
d
h
as
co
n
v
o
l
u
tio
n
n
e
u
r
al
n
et
w
o
r
k
s
(
C
NNs),
w
h
ic
h
ar
e
co
m
m
o
n
l
y
u
s
ed
to
class
i
f
y
o
b
j
ec
ts
[
1
1
]
.
C
NN
h
a
s
a
s
p
atial
s
tr
u
ctu
r
e
co
n
s
i
s
ti
n
g
o
f
la
y
e
r
s
o
f
s
ec
r
et
u
n
i
ts
[
1
2
]
,
in
clu
d
i
n
g
co
n
v
o
l
u
tio
n
,
p
o
o
lin
g
,
an
d
f
u
ll
y
co
n
n
ec
ted
la
y
er
s
.
I
n
t
h
e
co
n
v
o
l
u
ti
o
n
la
y
er
,
it
ac
ts
a
s
a
ch
ar
ac
ter
is
tic
s
ep
ar
ato
r
f
r
o
m
t
h
e
g
i
v
e
n
in
p
u
t
i
m
ag
e.
A
p
o
o
lin
g
la
y
er
p
r
o
ce
s
s
p
r
o
d
u
ce
s
d
o
w
n
-
s
a
m
p
l
in
g
f
o
r
w
ar
d
w
it
h
t
h
e
s
p
atial
ar
e
a.
F
in
all
y
,
t
h
e
f
u
ll
y
co
n
n
ec
t
ed
la
y
er
o
p
er
ate
s
as
a
clas
s
if
ier
t
h
at
p
r
ed
icts
th
e
p
r
o
d
u
ct
o
f
t
h
e
in
p
u
t
p
ict
u
r
e.
T
h
e
co
m
b
i
n
atio
n
o
f
t
h
e
s
e
la
y
er
s
e
n
ab
les
u
s
to
d
is
c
o
v
er
a
h
ier
ar
ch
ical
in
ter
p
r
etatio
n
o
f
t
h
e
in
p
u
t p
ict
u
r
e.
I
n
th
i
s
ac
t,
w
e
o
b
j
ec
tiv
e
a
t th
e
ap
p
r
o
p
r
ia
te
C
NN
to
i
n
cr
e
ase
t
h
e
ac
c
u
r
ac
y
o
f
th
e
b
r
ain
tr
an
s
cr
ip
tio
n
m
o
d
el
b
y
e
m
p
lo
y
i
n
g
f
ea
t
u
r
es
f
r
o
m
th
e
f
u
ll
y
co
n
n
ec
ted
la
y
er
.
Fo
r
th
is
r
ea
s
o
n
,
t
h
r
ee
ad
v
an
ce
d
le
v
els
o
f
C
NN
s
ar
e
u
s
ed
: V
GG1
6
,
R
esNe
t5
0
,
an
d
Xce
p
tio
n
,
to
lear
n
th
e
v
ec
to
r
p
r
o
p
er
ties
o
f
f
ig
u
r
es
co
n
ce
p
t
s
co
n
n
ec
ted
w
i
th
ea
c
h
n
o
tio
n
.
VGG1
6
[
1
3
]
T
h
e
im
p
o
r
tan
t
VGG1
6
s
tr
u
ctu
r
e
is
th
e
p
o
p
u
lar
C
NN
m
o
d
el.
T
h
e
n
etw
o
r
k
also
p
r
esen
ts
th
e
h
ig
h
est
5
test
ac
cu
r
ac
y
in
I
m
ag
eNe
t
w
ith
9
2
.
7
%
f
o
r
im
ag
e
class
if
icatio
n
.
T
h
is
m
o
d
el
is
s
h
o
w
n
u
tili
zin
g
lay
er
s
.
C
o
n
v
o
lu
tio
n
al
la
y
er
s
3
x
3
ar
e
p
ac
k
ed
o
n
to
p
o
f
ea
ch
lay
er
,
an
d
th
e
o
th
er
tw
o
lay
er
s
ar
e
co
n
n
ec
ted
b
y
ea
ch
n
o
d
e,
w
ith
4
,
0
9
6
co
n
n
ec
tio
n
s
s
h
o
w
n
in
Fig
u
r
e
2
.
VGG1
6
co
n
s
tr
u
ctio
n
.
Fig
u
r
e
2
.
VGG1
6
co
n
s
tr
u
ctio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
F
u
n
ctio
n
ma
g
n
etic
r
eso
n
a
n
ce
ima
g
in
g
-
b
a
s
ed
b
r
a
in
d
ec
o
d
i
n
g
w
ith
visu
a
l sema
n
tic
…
(
P
iya
w
a
t
S
a
en
g
p
etc
h
)
6685
R
esNet5
0
[
1
4
]
R
esNet
w
o
n
th
e
f
ir
s
t
p
r
i
ze
o
f
I
L
SVR
C
2
0
0
5
f
o
r
its
im
ag
e
class
if
icatio
n
,
s
h
o
w
n
in
Fig
u
r
e
3
.
Sk
ip
co
n
n
ec
tio
n
in
R
esNet5
0
.
T
h
e
R
esNet5
0
s
tr
u
ctu
r
e
co
n
s
is
ts
o
f
5
0
d
ee
p
co
n
v
er
s
io
n
lay
er
s
.
I
n
th
is
n
etw
o
r
k
,
th
er
e
ar
e
a
to
tal
o
f
1
6
r
em
ain
in
g
b
lo
ck
s
,
ea
ch
w
ith
b
lo
ck
s
.
T
h
er
e
ar
e
th
r
ee
-
lay
er
f
ee
d
lay
er
s
r
ea
d
y
.
Fig
u
r
e
3
.
Sk
ip
co
n
n
ec
tio
n
i
n
R
esNet5
0
Xce
p
tio
n
[
1
5
]
Xce
p
tio
n
b
y
Go
o
g
le
s
tan
d
s
f
o
r
th
e
Ultim
ate
v
er
s
io
n
o
f
I
n
ce
p
tio
n
.
W
ith
a
m
itig
ated
d
ep
th
w
is
e
d
etac
h
ab
le
co
n
v
o
lu
tio
n
,
it
is
ev
en
b
etter
th
an
I
n
ce
p
tio
n
-
v
3
[
1
6
]
(
also
b
y
Go
o
g
le,
1
s
t Ru
n
n
er
Up
in
I
L
SVR
C
2
0
1
5
)
f
o
r
b
o
th
I
m
ag
eNe
t
I
L
SVR
C
an
d
J
FT
d
atasets
.
T
h
e
Xce
p
tio
n
co
n
s
tr
u
ctio
n
h
as
3
6
co
n
v
o
lu
tio
n
al
lay
er
s
b
u
ild
in
g
th
e
f
ea
tu
r
e
ex
tr
ac
ti
o
n
b
ase
o
f
th
e
n
etw
o
r
k
an
d
f
u
lly
-
co
n
n
ec
ted
lay
er
s
b
ef
o
r
e
th
e
lo
g
is
tic
r
eg
r
ess
io
n
lay
er
.
T
h
e
3
6
co
n
v
o
lu
tio
n
al
lay
er
s
ar
e
d
ef
in
ed
in
to
1
4
s
ec
tio
n
s
,
all
o
f
w
h
ich
h
av
e
d
ir
ec
t
r
esid
u
al
lin
k
s
ar
o
u
n
d
th
em
,
ex
ce
p
t
f
o
r
th
e
f
ir
s
t
an
d
last
u
n
its
.
Fig
u
r
e
4
d
em
o
n
s
tr
ate
s
th
e
Xce
p
tio
n
co
n
s
tr
u
ctio
n
.
Fig
u
r
e
4
.
Xce
p
tio
n
ar
ch
itect
u
r
e
T
o
r
ec
o
v
er
th
e
ch
ar
ac
ter
is
tic,
th
e
p
ictu
r
e
w
e
u
s
e
VGG1
6
,
R
esNet5
0
,
an
d
Xce
p
tio
n
ca
n
d
id
ly
w
ith
im
p
o
r
tan
ce
tr
ain
ed
o
n
d
ata
s
ets
lar
g
er
as,
i.e
.
,
I
m
ag
eNe
t
h
as
co
m
b
in
ed
a
f
r
esh
lev
el
b
ef
o
r
e
p
r
o
d
u
ctio
n
o
r
lay
er
s
s
o
f
tm
ax
lay
er
.
A
ctio
n
s
as
a
s
ep
ar
ate
f
ea
tu
r
e
f
r
o
m
th
e
in
p
u
t
p
ictu
r
e.
T
h
e
n
eu
r
al
n
etw
o
r
k
w
as
r
etr
ain
ed
,
ad
o
p
tin
g
a
f
r
esh
s
et
o
f
p
ictu
r
es
w
h
ile
th
e
ea
r
lier
p
ar
t w
as
f
r
o
ze
n
.
F
(
x
)+
x
+
r
e
l
u
r
e
l
u
F
(
x
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Dec
em
b
er
2
0
2
0
:
6
6
8
2
-
6
6
9
0
6686
3
.
2
.
L
o
w
-
lev
el
v
is
ua
l descrip
t
o
rs
I
n
cr
ea
ti
n
g
an
i
m
a
g
e
r
ep
r
ese
n
t
atio
n
,
t
h
er
e
ar
e
m
a
n
y
p
r
ev
a
len
t
m
et
h
o
d
s
.
W
e
w
er
e
u
s
i
n
g
lo
w
-
le
v
el
d
escr
ip
to
r
s
to
ca
p
tu
r
e
g
en
er
al
r
ec
o
g
n
itio
n
f
ea
t
u
r
es
(
s
u
c
h
as
co
lo
r
,
s
u
r
f
ac
e,
s
tr
u
ctu
r
e,
ed
g
e)
.
T
h
ese
b
asic
f
ea
t
u
r
es
ca
n
b
e
d
r
a
w
n
i
m
m
e
d
iatel
y
f
r
o
m
t
h
e
i
m
a
g
e
a
n
d
e
asil
y
.
I
n
t
h
i
s
a
n
al
y
s
is
,
w
e
w
o
r
k
ed
o
n
t
h
e
v
is
u
al
p
r
o
p
er
ties
th
at
ar
e:
3
.
2
.
1
.
Si
m
ple c
o
lo
r
his
t
o
g
ra
m
(
CH
)
[
1
7
]
L
o
w
-
lev
el
f
ea
tu
r
e
u
tili
zin
g
a
co
lo
r
h
is
to
g
r
am
.
Set
to
th
e
d
is
tr
ib
u
tio
n
th
e
n
u
m
b
er
o
f
p
ix
els
in
th
e
im
ag
e
f
o
r
ea
ch
co
n
tain
er
.
W
e
w
ill
u
s
e
th
e
ef
f
ec
ts
o
f
th
e
co
lo
r
h
is
to
g
r
am
to
h
av
e
6
4
b
o
x
es
in
lin
e
w
ith
th
e
p
ar
t
o
f
th
e
co
lo
r
s
atu
r
atio
n
s
p
ec
tr
u
m
.
W
e
cr
ea
te
a
v
is
u
al
s
em
an
tic
m
o
d
el
w
ith
th
e
u
s
e
o
f
th
e
d
ef
au
lt
R
GB
co
lo
r
v
alu
es
an
d
s
et
th
e
n
u
m
b
er
o
f
ea
ch
co
m
p
o
n
en
t
o
f
th
e
R
GB
co
lo
r
in
to
f
o
u
r
,
w
h
ich
is
th
e
ea
s
iest
m
eth
o
d
.
(
4
x
4
x
4
).
3
.
2
.
2
.
E
dg
e
his
t
o
g
ra
m
(
EH
)
[
1
8
]
E
d
g
e
h
is
to
g
r
am
-
is
a
co
d
in
g
o
f
th
e
s
p
atial
d
is
tr
ib
u
tio
n
o
f
th
e
d
ir
ec
tio
n
ed
g
e.
Sp
ec
if
ically
,
th
e
im
ag
es
ar
e
d
iv
id
ed
in
to
72
b
o
x
es
,
ea
ch
o
f
w
h
ich
h
as
s
ev
er
al
co
r
n
er
s
w
ith
a
d
ir
ec
tio
n
th
at
is
m
ea
s
u
r
ed
in
5
-
d
eg
r
ee
in
ter
v
als.
I
n
t
h
is
ar
ticle,
u
s
e
C
an
n
y
f
ilter
s
f
o
r
ed
g
e
d
etec
tio
n
an
d
So
b
el
E
d
g
e
Dete
cto
r
o
p
er
ato
r
s
to
Me
asu
r
e
th
e
co
u
r
s
e
b
y
th
e
g
r
ad
ien
t
o
f
ev
er
y
ed
g
e
p
o
in
t.
T
h
e
s
em
an
tic
f
ea
tu
r
e
h
as
72
d
im
en
s
io
n
s
.
3
.
2
.
3
.
Co
lo
r
co
r
re
lo
g
ra
m
(
CO
RR
)
[
1
9
]
C
o
r
r
elo
g
r
am
co
lo
r
to
en
co
d
e
s
p
atial
r
elatio
n
s
h
ip
s
o
f
co
lo
r
s
.
On
e
o
f
th
e
tw
o
-
d
im
en
s
io
n
al
an
d
th
r
ee
-
d
im
en
s
io
n
al
h
is
to
g
r
am
is
th
e
co
lo
r
o
f
an
y
p
ix
el
an
d
th
r
ee
-
d
im
en
s
io
n
al
s
p
atial
d
is
tan
ce
s
.
T
h
e
co
lo
r
co
r
r
elo
g
r
am
is
d
ef
in
ed
as
:
,
(
)
=
1
∈
(
)
,
2
∈
[
2
∈
(
)
∨
|
1
−
2
|
=
]
(
2
)
w
h
er
e
|
1
−
2
|
is
th
e
m
ea
s
u
r
e
b
etw
ee
n
1
an
d
2
,
is
th
e
d
is
tan
ce
in
ter
v
als
n
u
m
b
er
,
,
∈
{
1
,
2
,
…
,
}
,
is
th
e
n
u
m
b
er
o
f
b
o
x
es
an
d
∈
{
1
,
2
,
…
,
}
,
.
W
e
d
is
tr
ib
u
te
th
e
R
GB
v
alu
e
elem
en
t
in
to
3
6
b
o
x
es
an
d
s
et
o
f
f
th
e
s
p
ac
e
m
etr
ic
to
4
o
d
d
in
ter
v
als
o
f
=
{
1
,
3
,
5
,
7
}
.
T
h
u
s
,
th
e
co
lo
r
co
r
r
elo
g
r
am
h
as
a
d
im
en
s
io
n
o
f
1
4
4
(
3
6
x
4
).
3
.
2
.
4
.
Sca
le
inv
a
ria
nce
f
ea
t
ure
t
ra
ns
f
o
r
m
(
SI
F
T
)
[
2
0
]
SIFT
is
th
e
m
o
s
t
u
s
ef
u
l
o
b
j
ec
t
id
en
tif
icatio
n
alg
o
r
ith
m
in
co
m
p
u
ter
v
is
io
n
(
C
V)
an
d
h
as
b
ee
n
u
s
ed
ex
ten
s
iv
ely
.
Fo
r
an
y
o
b
j
ec
t
in
th
e
p
ictu
r
e,
th
e
SIFT
k
ey
p
o
in
ts
o
n
p
u
r
p
o
s
e
ca
n
b
e
s
ep
ar
ated
in
to
s
m
aller
p
ar
ts
to
g
iv
e
a
s
p
ec
if
ic
d
escr
ip
to
r
th
at
d
ef
in
es
th
e
s
m
all
im
ag
e
ar
ea
ar
o
u
n
d
th
e
m
ar
k
o
n
th
at
o
b
j
ec
t
[
2
1
]
.
B
ec
au
s
e
o
f
th
e
SIFT
m
eth
o
d
r
esu
lts
in
th
e
ch
ar
ac
ter
is
tic
d
escr
ip
to
r
.
T
h
e
lar
g
e
s
ize
cr
ea
ted
in
th
e
p
ictu
r
e,
w
e
u
s
e
th
e
th
em
e
o
f
th
e
b
ag
o
f
f
ea
tu
r
es.
Fo
r
ea
ch
p
ictu
r
e,
f
ir
s
t
w
e
an
aly
ze
th
e
SIFT
d
escr
ip
to
r
ab
o
v
e
th
e
lo
ca
l
ar
ea
b
ased
o
n
th
e
k
ey
p
o
in
ts
.
W
e
th
en
ca
lcu
lated
th
e
v
ec
to
r
'
s
q
u
an
tity
o
n
th
e
SIFT
d
is
tr
ict
d
escr
ip
to
r
to
cr
ea
te
an
im
ag
e
v
o
ca
b
u
lar
y
u
s
in
g
k
-
m
ea
n
g
r
o
u
p
in
g
.
I
n
th
is
an
aly
s
is
,
w
e
cr
ea
ted
5
0
0
g
r
o
u
p
s
,
f
o
llo
w
in
g
in
th
e
s
ize
o
f
th
e
v
is
u
al
f
ea
tu
r
es
b
ein
g
5
0
0
,
f
o
r
r
ep
r
esen
tin
g
th
e
im
ag
e.
3
.
2
.
5
.
Wa
v
elet
T
ra
ns
f
o
r
m
(
WT
)
[
2
2
]
I
m
ag
e
r
ep
r
esen
tatio
n
R
em
ed
ies
s
u
r
f
ac
e
an
aly
s
is
o
f
th
e
im
ag
e
.
I
n
th
e
s
u
r
f
ac
e
an
aly
s
is
p
lan
,
th
e
w
av
elet
tr
an
s
f
o
r
m
ca
n
b
e
u
s
ed
to
d
etec
t
th
e
s
u
r
f
ac
e
o
f
th
e
im
ag
e
ef
f
ec
tiv
ely
.
W
av
elet
tr
an
s
f
o
r
m
s
d
o
n
e
o
n
p
ictu
r
es
th
at
s
u
g
g
est
r
ep
ea
ted
f
ilter
in
g
an
d
s
u
b
-
s
am
p
lin
g
.
A
t
ea
ch
lev
el,
p
ictu
r
es
ar
e
d
iv
id
ed
in
to
4
s
u
b
-
b
an
d
s
:
L
L
,
L
H,
HL
,
an
d
HH,
w
h
er
e
L
s
tan
d
s
f
o
r
lo
w
f
r
eq
u
en
cies,
an
d
H
m
ea
n
s
h
ig
h
f
r
eq
u
en
cies.
A
f
ter
th
at,
th
e
2
ty
p
es
o
f
w
av
elet
tr
an
s
f
o
r
m
s
ar
e
th
e
w
a
v
elet
tr
an
s
f
o
r
m
w
ith
th
e
p
y
r
am
id
s
tr
u
ctu
r
e
(
P
W
T
)
an
d
th
e
w
av
elet
tr
an
s
f
o
r
m
w
ith
th
e
tr
ee
s
tr
u
ctu
r
e
(
T
W
T
)
.
P
W
T
b
r
ea
k
s
th
e
L
L
b
an
d
r
ep
ea
ted
ly
,
an
d
T
W
T
b
r
ea
k
s
th
e
b
an
d
.
Oth
er
m
u
s
ic
to
p
r
eser
v
e
r
elev
an
t
in
f
o
r
m
atio
n
th
at
ap
p
ea
r
s
o
n
th
e
m
ed
iu
m
f
r
eq
u
en
cy
ch
an
n
el.
A
f
ter
p
ass
in
g
th
e
class
if
icatio
n
in
to
p
ar
ts
,
th
e
v
ec
to
r
is
ex
tr
ac
ted
u
s
in
g
th
e
m
ea
n
an
d
d
is
p
er
s
io
n
o
f
ea
ch
s
u
b
-
d
is
tr
ict
s
et
o
f
th
e
s
tr
en
g
th
d
is
tr
ib
u
tio
n
at
ea
ch
lev
el.
I
n
th
is
s
tu
d
y
,
w
e
u
s
ed
th
r
ee
lev
els
o
f
b
r
ea
k
d
o
w
n
s
to
cr
ea
te
v
ec
to
r
p
r
o
p
er
ties
o
f
ea
ch
o
b
j
ec
t
im
ag
e.
W
h
en
p
er
f
o
r
m
in
g
P
W
T
,
th
ey
r
ec
eiv
e
th
e
ch
ar
ac
ter
is
tics
o
f
a
2
4
(
3
x
4
x
2
)
v
ec
to
r
.
A
t
th
e
s
am
e
tim
e,
th
e
T
W
T
an
s
w
er
in
a
v
is
u
al
s
em
an
tic
f
ea
tu
r
e
s
ize
1
0
4
(
5
2
x
2
)
.
T
h
er
ef
o
r
e,
th
e
to
tal
w
av
elen
g
th
is
1
2
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
F
u
n
ctio
n
ma
g
n
etic
r
eso
n
a
n
ce
ima
g
in
g
-
b
a
s
ed
b
r
a
in
d
ec
o
d
i
n
g
w
ith
visu
a
l sema
n
tic
…
(
P
iya
w
a
t
S
a
en
g
p
etc
h
)
6687
4.
M
UL
T
I
T
A
SK
VIS
U
AL
SPA
CE
L
E
ARN
I
N
G
W
e
in
tr
o
d
u
ce
d
m
u
lti
-
task
lass
o
lear
n
in
g
(
MT
L
)
[
2
3
,
2
4
]
in
w
h
ich
m
an
y
lear
n
in
g
task
s
ar
e
im
p
r
o
v
ed
at
th
e
s
am
e
tim
e.
T
h
is
m
eth
o
d
p
r
o
d
u
ce
s
ac
cu
r
ate
p
r
ed
ictio
n
s
an
d
is
h
ig
h
ly
ef
f
ec
tiv
e.
I
n
class
if
icatio
n
,
MT
L
aim
s
to
in
cr
ea
s
e
th
e
ef
f
icien
cy
o
f
m
u
ltip
le
class
if
icatio
n
task
s
b
y
co
-
ope
r
atin
g
lear
n
in
g
.
On
e
ex
am
p
le
is
th
e
s
p
am
f
ilter
,
w
h
ich
ca
n
b
e
class
if
ied
as
a
co
m
p
lex
class
if
icatio
n
m
is
s
io
n
.
T
h
e
r
ea
s
o
n
f
o
r
th
is
is
th
at
lear
n
in
g
th
e
n
ativ
e
f
MRI
r
ep
r
esen
tatio
n
m
ay
n
o
t
h
av
e
en
o
u
g
h
im
p
ac
t
o
n
m
ar
k
in
g
m
u
ltip
le
class
es
d
u
e
to
th
e
t
r
ain
in
g
r
ep
r
esen
tatio
n
th
at
L
im
ited
an
d
ch
allen
g
in
g
to
f
in
d
d
ata
s
ets.
T
h
e
aim
is
to
lear
n
th
e
s
tan
d
ar
d
f
MRI
d
escr
ip
tio
n
s
r
elate
d
to
p
r
o
p
er
ties
th
at
ar
e
s
ep
ar
ated
f
r
o
m
m
an
y
class
es.
T
h
e
r
ea
s
o
n
f
o
r
th
is
is
th
at
lear
n
in
g
th
e
o
r
ig
in
al
f
MRI
r
ep
r
esen
tatio
n
m
ay
n
o
t
b
e
s
u
f
f
icien
t
f
o
r
m
an
y
s
tates'
ch
ar
ac
ter
izatio
n
d
u
e
to
th
e
s
m
all
an
d
lim
ited
tr
ain
in
g
ex
am
p
les.
A
s
ig
n
if
ican
t
im
p
ac
t
o
n
lear
n
in
g
is
th
e
f
in
d
in
g
o
f
r
ep
r
esen
tativ
es
o
f
lin
ea
r
an
d
n
o
n
lin
ea
r
tr
an
s
f
o
r
m
atio
n
s
o
f
f
MRI
im
ag
e
d
ata,
as
illu
s
tr
ated
b
y
m
ath
em
atica
l
eq
u
atio
n
s
.
C
o
n
s
id
er
=
{
}
=
1
is
a
s
et
o
f
s
ep
ar
ate
task
s
w
h
er
e
a
task
d
ir
ec
ts
o
n
lear
n
in
g
m
o
d
el
f
o
r
esti
m
ate
th
e
th
v
alu
e
.
T
r
ain
in
g
d
ataset
is
=
{
,
}
=
1
,
w
h
er
e
∈
is
th
e
th
t
r
ain
in
g
f
MRI
im
ag
e
m
ad
e
o
f
v
o
x
els
in
an
d
is
its
th
f
ea
tu
r
e
v
alu
e
.
T
h
e
tr
ain
in
g
f
MRI
d
ata
m
atr
ix
is
f
o
r
,
i
.
e
.
=
(
1
,
…
,
)
.
Fo
r
th
e
t
a
s
k
,
w
e
l
e
a
r
n
^
=
∑
+
=
1
w
h
e
r
e
^
,
,
∈
M
o
r
e
o
v
e
r
,
i
s
t
h
e
n
u
m
b
e
r
o
f
v
o
x
e
l
s
.
W
e
p
r
e
s
u
m
e
th
at
th
e
d
ata
is
r
eg
u
lated
s
o
th
e
co
n
s
tan
t
ter
m
s
ca
n
b
e
d
r
o
p
p
ed
,
i
.
e
.
an
d
^
h
av
e
av
er
ag
e
0
an
d
‖
‖
2
=
1
w
h
er
e
‖
.
‖
2
i
s
th
e
L
2
E
u
clid
ea
n
n
o
r
m
.
L
et
=
(
1
,
…
,
)
b
e
th
e
v
ec
to
r
o
f
all
co
ef
f
icien
ts
f
o
r
th
e
th
v
o
x
el
ac
r
o
s
s
v
ar
io
u
s
task
s
.
T
o
ac
h
iev
e
a
co
m
p
ac
t
an
d
d
is
cr
im
in
ativ
e
r
ep
r
esen
tatio
n
,
th
e
m
u
lti
-
task
L
ass
o
is
f
o
r
m
u
lated
as
th
e
an
s
w
er
to
th
e
o
p
tim
i
za
tio
n
p
r
o
b
lem
.
{
1
2
∑
‖
−
∑
=
1
‖
2
2
+
∑
‖
‖
∞
=
1
=
1
}
(
3
)
W
h
er
e
‖
‖
∞
=
|
|
is
th
e
s
u
p
-
n
o
r
m
in
th
e
E
u
clid
ea
n
s
p
ac
e
.
I
t
ad
d
itio
n
all
y
h
as
th
e
i
m
p
r
es
s
io
n
o
f
"
g
r
o
u
p
in
g
"
th
e
p
ar
ts
in
s
u
ch
th
at
t
h
e
y
ca
n
o
b
tain
ze
r
o
s
co
n
cu
r
r
en
tl
y
.
Af
ter
teac
h
in
g
m
o
d
el
s
,
w
e
s
till
h
a
v
e
to
b
u
ild
a
d
ec
is
io
n
r
u
le
to
ch
o
o
s
e
th
e
m
o
s
t
i
m
p
o
r
ta
n
t
clas
s
to
b
e
r
elate
d
to
a
g
iv
e
n
f
MRI
i
m
ag
e
.
Fo
r
a
g
iv
en
f
M
R
I
i
m
ag
e
,
p
r
ed
icted
f
ea
tu
r
e
v
alu
e
s
ar
e
o
b
tain
ed
b
y
u
s
i
n
g
co
ef
f
icie
n
t
s
.
L
ast
l
y
,
P
ea
r
s
o
n
's
co
r
r
elatio
n
co
ef
f
icie
n
t
[
2
5
]
ar
e
u
s
ed
to
m
ea
s
u
r
e
t
h
e
as
s
o
ciatio
n
b
et
wee
n
X
a
n
d
a
tar
g
et
class
C
.
(
,
)
=
∑
(
(
)
−
´
)
(
−
´
)
=
1
√
∑
(
−
´
)
2
=
1
×
√
∑
(
−
´
)
2
=
1
(
4
)
w
h
er
e
m
ap
p
i
n
g
f
u
n
c
tio
n
i
s
(
)
th
at
r
ec
o
n
s
tr
u
ct
s
X
to
t
h
e
th
f
ea
t
u
r
e
v
al
u
e
.
T
h
e
r
esu
lt
w
it
h
th
e
h
ig
h
es
t
co
r
r
elatio
n
f
MRI
co
n
ce
p
t
.
5.
E
XP
E
R
I
M
E
NT
S
E
T
UP
5
.
1
.
Da
t
a
s
et
s
I
n
th
i
s
r
esear
c
h
,
we
u
s
e
f
MRI
3
D
d
ata
f
r
o
m
C
ar
n
e
g
i
e
Me
ll
o
n
U
n
i
v
er
s
it
y
,
co
llect
ed
f
r
o
m
9
r
ig
h
t
-
h
an
d
ed
ad
u
lt
v
o
l
u
n
teer
s
,
co
n
s
is
t
in
g
o
f
f
M
R
I
d
ata
th
at
allo
w
s
v
o
l
u
n
teer
s
to
v
ie
w
6
0
lin
e
d
r
a
w
in
g
s
a
n
d
n
o
u
n
s
,
s
i
x
ti
m
es
ea
c
h
.
I
m
a
g
es
(
6
X6
0
=
3
6
0
)
1
f
MRI
i
m
ag
e
w
ill
b
e
co
n
v
er
ted
to
a
v
o
x
e
l
v
ec
to
r
w
it
h
ap
p
r
o
x
im
a
tel
y
2
0
0
0
0
v
o
x
els.
T
h
en
d
ec
r
ea
s
e
th
e
s
ize
o
f
th
e
v
ec
to
r
to
5
0
0
v
o
x
els
u
s
i
n
g
t
h
e
s
ea
r
ch
l
ig
h
t
m
et
h
o
d
[
2
4
]
.
I
n
th
e
tech
n
iq
u
e
[
6
]
an
d
[
7
]
,
f
MRI
d
ata
ar
e
d
ec
r
ea
s
ed
to
5
0
0
-
f
ea
t
u
r
es.
5
.
2
.
T
ex
t
-
ba
s
ed
c
f
ea
t
ures
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
p
er
f
o
r
m
i
n
g
i
s
co
m
p
ar
ed
w
it
h
t
h
e
tex
t
-
b
ased
s
e
m
an
tic
f
ea
t
u
r
es
s
tate
-
of
-
t
h
e
-
ar
t
m
o
d
el
s
,
as
d
escr
ib
ed
f
o
llo
w
s
:
Ver
b
2
5
[
6
]
p
r
o
v
id
es
th
e
n
o
tio
n
i
n
th
e
f
o
r
m
o
f
t
h
e
n
o
u
n
an
d
d
escr
ib
in
g
t
h
e
co
-
o
cc
u
r
r
en
ce
v
ec
to
r
b
et
w
ee
n
n
o
u
n
s
a
n
d
v
er
b
s
o
f
2
5
n
a
m
e
s
s
u
c
h
as
"
r
u
n
"
,
"
p
u
s
h
"
,
"
ea
t
"
an
d
m
a
n
y
o
th
er
s
.
T
h
ese
co
m
m
o
n
v
er
b
s
ar
e
o
f
ten
d
ef
i
n
ite
n
o
u
n
s
i
n
E
n
g
l
is
h
s
e
n
te
n
ce
s
an
d
d
esi
g
n
ed
b
y
t
h
e
s
tr
u
ct
u
r
alis
t
s
.
I
n
an
i
n
v
e
s
ti
g
atio
n
,
Ver
b
-
25
is
a
p
r
ac
tical
ef
f
ec
t o
f
th
is
d
ata
s
et
.
Hu
m
an
2
1
8
[
7
]
h
o
ld
s
th
e
n
o
tio
n
in
th
e
f
o
r
m
o
f
2
1
8
attr
ib
u
tes
.
T
h
e
p
atter
n
v
ec
to
r
is
o
b
tain
ed
by
2
1
8
q
u
esti
o
n
s
,
s
u
c
h
as
"
I
s
it
co
l
d
?
",
"
I
s
it
h
o
t
?
"
,
"
C
a
n
yo
u
w
a
lk
o
n
it?
"
e
tc
.
T
h
e
lin
g
u
is
ts
cr
ea
ted
th
ese
q
u
esti
o
n
s
a
n
d
co
llected
t
h
e
a
n
s
w
er
s
f
r
o
m
clo
u
d
s
o
u
r
cin
g
a
n
d
co
m
p
u
ted
t
h
e
a
v
er
ag
e
an
s
wer
s
co
r
r
elate
d
to
ea
ch
n
o
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Dec
em
b
er
2
0
2
0
:
6
6
8
2
-
6
6
9
0
6688
5
.
3
.
I
m
a
g
e
-
ba
s
ed
f
ea
t
ures
I
n
th
i
s
r
esear
ch
,
w
e
h
a
v
e
s
ele
cted
v
is
u
a
l
p
r
o
p
er
ties
f
r
o
m
p
i
ctu
r
es
f
o
r
th
e
d
esi
g
n
o
f
ex
p
er
i
m
e
n
ts
b
y
u
s
i
n
g
p
r
o
p
er
ties
f
r
o
m
i
m
a
g
es
f
r
o
m
lo
w
er
to
u
p
p
er
lev
els.
C
o
lo
r
h
is
to
g
r
a
m
(
C
H)
,
co
lo
r
co
r
r
elo
g
r
a
m
(
C
OR
R
)
,
ed
g
e
h
i
s
to
g
r
a
m
(
E
H)
,
W
av
elet
T
r
an
s
f
o
r
m
(
W
T
)
,
B
o
W
+
SIFT
ar
e
o
n
NUS
-
W
I
DE
[
2
6
]
d
atab
ase
r
esear
ch
.
T
h
e
NUS
-
W
I
DE
d
ataset
h
o
l
d
s
2
6
9
,
6
4
8
p
ictu
r
es
co
m
p
i
le
d
f
r
o
m
th
e
Fl
ick
r
o
n
li
n
e
p
h
o
to
g
r
ap
h
d
atab
ase
to
ad
ap
t
to
th
e
f
MRI
i
m
a
g
e
co
n
ce
p
t
an
d
Use
th
e
f
o
llo
w
i
n
g
ch
ar
ac
ter
is
tic
s
f
r
o
m
t
h
e
C
N
Ns
,
n
a
m
el
y
VGG1
6
,
R
esNet5
0
,
a
n
d
Xce
p
tio
n
,
a
n
d
u
s
e
t
h
e
p
r
o
p
er
ties
f
r
o
m
all
t
h
e
i
m
a
g
es
to
cr
ea
te
a
f
ea
t
u
r
e
v
ec
to
r
to
d
escr
ib
e
th
e
co
n
ce
p
t.
6.
RE
SU
L
T
S
AND
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
is
a
test
to
co
m
p
a
r
e
th
e
ef
f
icie
n
c
y
o
f
u
s
i
n
g
th
e
p
r
o
p
er
ties
o
b
tain
ed
f
r
o
m
te
x
t
u
s
i
n
g
n
i
n
e
f
MRI
i
m
a
g
es
o
f
ea
c
h
v
o
lu
n
te
er
,
3
6
0
im
a
g
es
f
r
o
m
6
0
i
m
ag
es
d
iv
id
ed
b
y
co
n
ce
p
t,
f
o
r
u
s
e
in
teac
h
i
n
g
3
0
0
i
m
a
g
es
a
n
d
6
0
i
m
ag
e
s
f
o
r
tes
tin
g
to
f
i
n
d
t
h
e
ef
f
ec
ti
v
e
n
es
s
o
f
m
o
d
els
a
n
d
f
ea
t
u
r
es
e
x
tr
ac
ted
w
it
h
d
i
f
f
er
e
n
t
m
et
h
o
d
s
.
T
a
b
le
1
s
h
o
w
s
p
r
e
d
ictab
ilit
y
o
f
t
h
e
co
n
ce
p
t
s
ee
n
i
n
t
h
e
L
R
m
o
d
el
a
cc
u
r
ac
y
(
%)
f
r
o
m
9
f
MRI
v
o
lu
n
teer
s
an
d
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
s
(
i
.
e
.
,
W
T
,
E
DH,
R
esNet5
0
,
an
d
Xce
p
tio
n
)
.
I
t
ca
n
b
e
o
b
s
er
v
ed
th
at
th
e
p
r
o
p
o
s
ed
v
is
u
al
f
ea
tu
r
e
s
ig
n
if
ican
t
l
y
e
n
h
a
n
ce
s
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
b
r
ain
d
ec
o
d
in
g
.
Mo
r
eo
v
er
,
th
e
b
est
p
er
f
o
r
m
a
n
ce
m
o
d
els
ar
e
o
b
ta
in
ed
b
y
u
s
i
n
g
ex
tr
ac
ted
f
ea
t
u
r
es
b
ased
o
n
VGG1
6
,
R
esNet5
0
,
an
d
Xce
p
tio
n
.
T
h
e
m
u
lti
-
ta
s
k
lear
n
in
g
(
MT
L
)
m
et
h
o
d
is
also
co
m
p
ar
ed
w
i
th
a
lin
ea
r
r
eg
r
es
s
io
n
(
LR
)
m
e
th
o
d
.
A
s
s
h
o
w
n
in
T
ab
le
1
an
d
T
ab
le
2
all
th
e
MT
L
m
o
d
els
s
ig
n
i
f
ica
n
tl
y
o
u
t
p
er
f
o
r
m
ed
t
h
e
L
R
m
o
d
el
s
.
T
h
u
s
,
t
h
e
g
i
v
en
r
e
s
u
l
t
s
e
m
p
h
a
s
ize
th
e
h
i
g
h
l
ig
h
t
o
f
t
h
e
MT
L
ap
p
r
o
ac
h
f
o
r
i
m
p
r
o
v
in
g
th
e
g
e
n
er
aliza
tio
n
p
er
f
o
r
m
an
ce
.
T
h
e
leav
e
-
t
w
o
-
o
u
t c
r
o
s
s
-
v
alid
atio
n
tech
n
iq
u
e
[
6
,
7
]
h
av
e
b
ee
n
e
m
p
lo
y
ed
f
o
r
ev
alu
ati
n
g
th
e
e
f
f
ec
ti
v
e
n
ess
o
f
th
e
v
is
u
al
f
ea
t
u
r
e
f
o
r
d
ec
o
d
in
g
n
o
v
el
co
n
ce
p
ts
(
u
n
s
ee
n
)
.
T
ab
le
1
.
P
r
ed
ictab
ilit
y
o
f
t
h
e
c
o
n
ce
p
t seen
i
n
th
e
L
R
m
o
d
el
A
cc
u
r
ac
y
(
%)
f
r
o
m
9
f
M
R
I
v
o
lu
n
teer
s
M
o
d
e
l
v
o
l
u
n
t
e
e
r
A
v
g
.
#
F
e
a
t
u
r
e
1
2
3
4
5
6
7
8
9
V
e
r
b
s 2
5
68
.
99
62
.
45
64
.
85
63
.
41
54
.
74
58
.
57
55
.
26
55
.
05
59
.
98
60
.
37
25
H
u
ma
n
2
1
8
62
.
14
58
.
43
54
.
90
58
.
45
56
.
30
53
.
30
56
.
52
48
.
59
49
.
53
55
.
35
2
1
8
CH
59
.
92
60
.
03
59
.
26
56
.
51
50
.
08
50
.
46
54
.
65
51
.
06
51
.
11
54
.
79
64
C
O
R
R
61
.
47
59
.
40
53
.
19
54
.
92
49
.
76
50
.
07
54
.
58
58
.
53
54
.
85
55
.
20
1
4
4
WT
61
.
23
57
.
38
55
.
77
54
.
78
50
.
41
52
.
46
51
.
61
53
.
05
56
.
58
54
.
81
1
2
8
ED
H
60
.
27
55
.
53
56
.
33
55
.
64
53
.
40
51
.
25
51
.
95
53
.
29
51
.
05
54
.
30
73
B
o
W
(
S
I
F
T
)
70
.
52
63
.
03
61
.
58
61
.
55
49
.
10
53
.
82
54
.
61
52
.
61
55
.
81
60
.
07
5
0
0
V
G
G
1
6
65
.
21
60
.
79
59
.
29
62
.
31
54
.
60
56
.
51
53
.
80
54
.
13
51
.
91
57
.
62
4
,
0
9
6
R
e
sN
e
t
5
0
77
.
15
69
.
27
67
.
92
71
.
09
60
.
60
62
.
80
63
.
45
54
.
29
59
.
63
65
.
13
2
,
0
4
8
X
c
e
p
t
i
o
n
56
.
76
62
.
67
55
.
39
56
.
12
53
.
22
54
.
49
53
.
50
52
.
28
54
.
56
55
.
33
2
,
0
4
8
W
e
test
ed
th
e
m
o
d
el
o
n
all
th
e
f
M
R
I
i
m
a
g
es
o
f
5
8
co
n
ce
p
ts
f
o
r
th
e
tr
ain
i
n
g
s
e
t,
an
d
th
en
f
MRI
i
m
a
g
es
o
f
2
co
n
ce
p
ts
f
o
r
th
e
test
in
g
s
et
.
T
h
e
c
o
m
p
ar
is
o
n
r
es
u
lts
o
f
all
th
e
p
r
ed
ictin
g
2
u
n
s
ee
n
co
n
ce
p
ts
m
o
d
el
h
av
e
b
ee
n
il
lu
s
tr
ated
in
T
ab
le
2
.
As
r
ec
o
r
d
ed
in
T
ab
le
2
,
B
o
th
VGG1
6
a
n
d
R
es
Net5
0
s
i
g
n
i
f
ica
n
tl
y
b
etter
ed
Ver
b
2
5
an
d
Hu
m
a
n
2
1
8
,
w
h
e
r
ea
s
th
e
R
es
Net5
0
u
s
ed
1
0
6
v
o
x
el
s
o
f
th
e
w
h
o
le
b
r
ain
s
h
o
w
ed
th
e
h
i
g
h
es
t
co
r
r
ec
tn
ess
.
Fi
g
u
r
e
5
d
e
m
o
n
s
tr
ates
t
h
at
t
h
e
MT
L
m
o
d
el
e
x
ce
ed
s
t
h
e
L
R
m
o
d
el
f
o
r
d
e
co
d
in
g
V1
a
n
d
V2
o
v
er
all
th
e
s
e
m
a
n
tic
f
ea
tu
r
e
s
.
As
a
r
esu
lt,
t
h
e
h
i
g
h
-
lev
e
l
v
i
s
u
al
f
ea
tu
r
e
is
b
etter
f
o
r
id
en
ti
f
y
in
g
b
r
ain
ac
ti
v
it
y
p
atter
n
s
in
ter
m
s
o
f
ac
cu
r
ac
y
.
T
ab
le
2
.
P
r
ed
ictab
ilit
y
o
f
t
h
e
c
o
n
ce
p
t seen
i
n
th
e
MT
L
m
o
d
e
l A
cc
u
r
ac
y
(
%)
f
r
o
m
9
f
M
R
I
v
o
lu
n
teer
s
M
o
d
e
l
v
o
l
u
n
t
e
e
r
s
A
v
g
.
#
F
e
a
t
u
r
e
S
e
l
e
c
t
e
d
V
o
x
e
l
s
1
2
3
4
5
6
7
8
9
V
e
r
b
s
25
70
.
16
67
.
08
67
.
68
66
.
35
59
.
03
56
.
56
58
.
23
56
.
65
57
.
14
62
.
10
25
3
2
8
H
u
ma
n
2
1
8
61
.
38
60
.
82
57
.
94
59
.
94
57
.
59
55
.
22
58
.
13
49
.
66
49
.
46
56
.
68
2
1
8
4
4
4
CH
64
.
79
63
.
47
60
.
87
58
.
45
53
.
11
52
.
25
58
.
95
50
.
78
51
.
01
57
.
08
64
1
4
7
C
O
R
R
63
.
38
63
.
12
58
.
64
58
.
73
51
.
99
52
.
07
58
.
50
55
.
35
57
.
34
57
.
68
1
4
4
2
3
3
WT
59
.
36
61
.
79
59
.
02
61
.
23
52
.
32
53
.
85
57
.
61
53
.
85
53
.
42
56
.
94
1
2
8
1
2
7
ED
H
67
.
04
57
.
25
58
.
47
59
.
55
54
.
68
51
.
32
54
.
89
54
.
16
53
.
05
56
.
71
73
1
3
4
B
o
W
(
S
I
F
T
)
79
.
50
71
.
20
70
.
78
69
.
41
53
.
87
57
.
59
59
.
47
55
.
85
55
.
59
63
.
70
5
0
0
2
4
1
VGG
16
79
.
57
70
.
04
70
.
71
71
.
86
61
.
58
65
.
47
64
.
59
54
.
13
58
.
20
66
.
24
4
,
0
9
6
4
8
5
R
e
sN
e
t
50
79
.
18
72
.
36
71
.
12
73
.
37
62
.
52
64
.
73
63
.
90
55
.
30
58
.
43
66
.
77
2
,
0
4
8
4
8
2
X
c
e
p
t
i
o
n
82
.
72
87
.
67
79
.
29
86
.
36
77
.
27
87
.
01
78
.
78
82
.
32
78
.
57
82
.
22
2
,
0
4
8
4
3
5
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
F
u
n
ctio
n
ma
g
n
etic
r
eso
n
a
n
ce
ima
g
in
g
-
b
a
s
ed
b
r
a
in
d
ec
o
d
i
n
g
w
ith
visu
a
l sema
n
tic
…
(
P
iya
w
a
t
S
a
en
g
p
etc
h
)
6689
(
a)
(
b
)
Fig
u
r
e
5
.
P
er
f
o
r
m
a
n
ce
co
m
p
ar
is
o
n
o
f
M
T
L
m
o
d
e
l
a
n
d
L
R
m
o
d
e
l
o
n
V1
a
n
d
V
2
7.
CO
NCLU
SI
O
N
I
n
th
i
s
p
ap
er
,
w
e
h
av
e
p
r
o
p
o
s
ed
an
e
x
o
tic
f
M
R
I
b
r
ain
tr
a
n
s
cr
ip
tio
n
m
o
d
el
u
s
i
n
g
i
m
a
g
e
-
to
-
i
m
a
g
e
f
ea
t
u
r
es.
T
h
e
v
is
u
al
c
h
ar
ac
ter
is
tics
o
f
m
a
n
y
ap
p
lican
t
s
ar
e
ex
a
m
in
ed
f
r
o
m
lo
w
to
h
i
g
h
-
lev
el
f
ea
tu
r
e
s
th
a
t
f
o
cu
s
o
n
th
e
f
M
R
I
d
ec
o
d
in
g
m
o
d
el
p
r
ec
is
el
y
.
A
l
s
o
,
th
e
s
tu
d
y
p
r
esen
ted
is
a
co
m
p
ac
t
an
d
d
is
cr
i
m
i
n
ati
n
g
p
atter
n
lear
n
in
g
-
th
e
r
elatio
n
s
h
i
p
b
et
w
ee
n
t
h
e
v
o
x
e
l
ac
ti
v
at
io
n
m
o
d
el
an
d
t
h
e
i
m
a
g
e
p
r
o
p
er
ties
to
d
ec
ip
h
er
th
e
co
n
ce
p
t.
T
h
e
r
es
u
lt
s
s
h
o
w
th
e
o
p
er
atio
n
al
s
u
cc
e
s
s
o
f
o
u
r
m
o
d
el
a
n
d
t
h
e
ad
v
a
n
ta
g
e
o
v
er
s
tate
-
of
-
t
h
e
-
ar
t
b
r
ain
d
ec
o
d
in
g
w
it
h
f
M
R
I
.
RE
F
E
R
E
NC
E
S
[1
]
R.
S
it
a
ra
m
,
e
t
a
l.
,
“
f
M
RI
b
ra
in
-
c
o
m
p
u
ter i
n
terf
a
c
e
s,”
IEE
E
S
ig
n
a
l
Pro
c
e
ss
.
M
a
g
.
,
v
o
l
.
2
5
,
n
o
.
1
,
p
p
.
9
5
-
1
0
6
,
2
0
0
8
.
[2
]
Y.
-
F
.
L
a
i,
M
.
-
Y.
C
h
e
n
,
a
n
d
H.
-
S
.
Ch
ian
g
,
“
Co
n
stru
c
ti
n
g
th
e
l
ie
d
e
tec
ti
o
n
sy
ste
m
w
it
h
f
u
z
z
y
r
e
a
so
n
in
g
a
p
p
r
o
a
c
h
,
”
Gr
a
n
u
l
.
Co
m
p
u
t.
,
v
o
l.
3
,
n
o
.
2
,
p
p
.
1
6
9
-
1
7
6
,
2
0
1
7
.
[3
]
J.
K.
Hu
tt
u
n
e
n
e
t
a
l
.
,
“
De
tec
ti
o
n
o
f
H
y
p
e
re
x
c
it
a
b
il
it
y
b
y
F
u
n
c
ti
o
n
a
l
M
a
g
n
e
ti
c
Re
so
n
a
n
c
e
Im
a
g
in
g
a
f
ter
Ex
p
e
rime
n
tal
T
ra
u
m
a
ti
c
Bra
in
In
j
u
ry
,
”
J
o
u
rn
a
l
o
f
n
e
u
ro
tr
a
u
ma
,
v
o
l.
3
5
,
n
o
.
2
2
,
p
p
.
2
7
0
8
-
2
7
1
7
,
2
0
1
8
.
[4
]
K.
J.
W
o
rsle
y
e
t
a
l.
,
“
A
G
e
n
e
ra
l
S
tatisti
c
a
l
A
n
a
l
y
sis
f
o
r
f
M
RI
Da
ta,
”
Ne
u
ro
ima
g
e
,
v
o
l.
1
5
,
n
o
.
1
,
p
p
.
1
-
1
5
,
2
0
0
2
.
[5
]
A
.
M
a
h
m
o
u
d
i,
e
t
a
l.
,
“
M
u
lt
iv
o
x
e
l
p
a
tt
e
rn
a
n
a
ly
si
s
f
o
r
f
M
RI
d
a
ta:
A
re
v
i
e
w
,
”
Co
mp
u
t.
M
a
th
.
M
e
th
o
d
s
M
e
d
.
,
v
o
l.
2
0
1
2
,
2
0
1
2
.
[6
]
T
.
M
.
M
it
c
h
e
ll
e
t
a
l.
,
“
P
re
d
icti
n
g
h
u
m
a
n
b
ra
in
a
c
ti
v
it
y
a
ss
o
c
iate
d
w
it
h
th
e
m
e
a
n
in
g
s
o
f
n
o
u
n
s.,
”
sc
ien
c
e
3
2
0
.
5
8
8
0
,
v
o
l.
3
2
0
,
n
o
.
5
8
8
0
,
p
p
.
1
1
9
1
-
1
1
9
5
,
2
0
0
8
.
[7
]
M
.
P
a
latu
c
c
i,
e
t
a
l.
,
“
Zero
-
S
h
o
t
L
e
a
rn
in
g
w
it
h
S
e
m
a
n
ti
c
Ou
tp
u
t
C
o
d
e
s,”
Ne
u
ra
l
I
n
f.
Pr
o
c
e
ss
.
S
y
st.
,
p
p
.
1
-
9
,
2
0
0
9
.
[8
]
L
.
P
i
p
a
n
m
a
e
k
a
p
o
rn
,
e
t
a
l.
,
“
De
sig
n
in
g
S
e
m
a
n
ti
c
F
e
a
tu
re
S
p
a
c
e
s f
o
r
Bra
in
-
Re
a
d
in
g
,
”
Esa
n
n
,
p
p
.
2
2
-
2
4
,
2
0
1
5
.
[9
]
A
.
B.
Je
lo
d
a
r,
e
t
a
l.
,
“
W
o
rd
Ne
t
Ba
se
d
F
e
a
tu
re
s
f
o
r
P
re
d
icti
n
g
Br
a
in
A
c
ti
v
it
y
a
ss
o
c
iate
d
w
it
h
m
e
a
n
in
g
s
o
f
n
o
u
n
s,”
NAA
CL
HLT
2
0
1
0
Fi
rs
t
W
o
rk
.
C
o
mp
u
t
.
Ne
u
ro
li
n
g
u
ist
ics
,
p
p
.
1
8
-
2
6
,
2
0
1
0
.
[1
0
]
P
.
S
a
e
n
g
p
e
tch
,
L
.
P
ip
a
n
m
e
m
e
k
a
p
o
r
n
,
a
n
d
S
.
Ka
m
o
lsa
n
ti
ro
j,
“
Vi
su
a
l
re
p
re
se
n
tatio
n
m
o
d
e
l
f
o
r
fM
RI
-
b
a
se
d
b
ra
in
d
e
c
o
d
i
n
g
,
”
ACM
In
t.
C
o
n
f
.
Pro
c
e
e
d
in
g
S
e
r.
,
p
p
.
5
8
-
6
3
,
2
0
1
9
.
[1
1
]
K.
Ka
v
u
k
c
u
o
g
lu
,
e
t
a
l.
,
“
Lea
rn
in
g
c
o
n
v
o
lu
ti
o
n
a
l
f
e
a
tu
re
h
iera
rc
h
ies
f
o
r
v
isu
a
l
re
c
o
g
n
it
io
n
,
”
Ad
v
.
Ne
u
ra
l
In
f.
Pro
c
e
ss
.
S
y
st.
2
3
2
4
t
h
A
n
n
u
.
Co
n
f
.
Ne
u
ra
l
I
n
f.
Pro
c
e
ss
.
S
y
st.
2
0
1
0
,
n
o
.
1
,
p
p
.
1
-
9
,
2
0
1
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Dec
em
b
er
2
0
2
0
:
6
6
8
2
-
6
6
9
0
6690
[1
2
]
C.
S
a
m
p
le et al.
,
“
A
g
e
n
e
ra
l
m
o
d
e
li
n
g
f
ra
m
e
w
o
rk
f
o
r
d
e
sc
rib
in
g
sp
a
ti
a
ll
y
stru
c
tu
re
d
p
o
p
u
lati
o
n
d
y
n
a
m
ic
s,”
Eco
lo
g
y
a
n
d
Evo
l
u
ti
o
n
,
v
o
l.
8
,
n
o
.
1
,
p
p
.
4
9
3
-
5
0
8
,
2
0
1
8
.
[1
3
]
K.
S
im
o
n
y
a
n
a
n
d
A
.
Zi
ss
e
r
m
a
n
,
“
V
e
ry
d
e
e
p
c
o
n
v
o
lu
ti
o
n
a
l
n
e
tw
o
rk
s
f
o
r
larg
e
-
s
c
a
le
i
m
a
g
e
re
c
o
g
n
it
io
n
,
”
a
rXiv
p
re
p
rin
t
a
rXiv:1
4
0
9
.
1
5
5
6
,
2
0
1
4
.
[
1
4
]
S
.
W
u
,
S
.
Z
h
o
n
g
,
a
n
d
Y
.
L
i
u
,
“
D
e
e
p
r
e
s
i
d
u
a
l
l
e
a
r
n
i
n
g
f
o
r
i
m
a
g
e
s
t
e
g
a
n
a
l
y
s
i
s
,
”
M
u
l
t
i
m
e
d
.
T
o
o
l
s
A
p
p
l
.
,
p
p
.
1
-
1
7
,
2
0
1
7
.
[1
5
]
F
.
Ch
o
ll
e
t
,
“
X
c
e
p
ti
o
n
:
De
e
p
Lea
rn
in
g
w
it
h
D
e
p
th
w
ise
S
e
p
a
ra
b
le
Co
n
v
o
l
u
ti
o
n
s,”
Pro
c
e
e
d
in
g
s
o
f
th
e
IEE
E
c
o
n
fer
e
n
c
e
o
n
c
o
mp
u
ter
v
isio
n
a
n
d
p
a
tt
e
rn
re
c
o
g
n
it
io
n
,
p
p
.
1
2
5
1
-
1
2
5
8
,
2
0
1
7
.
[1
6
]
C.
S
z
e
g
e
d
y
,
e
t
a
l.
,
“
R
e
th
in
k
in
g
th
e
In
c
e
p
ti
o
n
A
rc
h
it
e
c
tu
re
f
o
r
Co
m
p
u
ter
V
isio
n
,
”
Pro
c
.
IEE
E
Co
mp
u
t.
S
o
c
.
Co
n
f
.
Co
mp
u
t
.
Vi
s.
Pa
t
ter
n
Rec
o
g
n
it
.
,
v
o
l.
2
0
1
6
,
p
p
.
2
8
1
8
-
2
8
2
6
,
2
0
1
5
.
[1
7
]
L
.
S
h
a
p
iro
a
n
d
G
.
S
to
c
k
m
a
n
,
"
Co
m
p
u
ter V
isio
n
,
"
Pre
n
ti
c
e
Ha
ll
,
v
o
l.
2
1
,
n
o
.
3
.
2
0
0
1
.
[1
8
]
N.
P
ra
jap
a
ti
,
A
.
K.
Na
n
d
a
n
w
a
r,
a
n
d
G
.
S
.
P
ra
jap
a
ti
,
“
e
d
g
e
h
isto
g
ra
m
d
e
sc
rip
to
r,
g
e
o
m
e
tri
c
m
o
m
e
n
t
a
n
d
so
b
e
l
e
d
g
e
d
e
tec
to
r
c
o
m
b
in
e
d
f
e
a
tu
re
s
b
a
se
d
o
b
jec
t
re
c
o
g
n
it
io
n
a
n
d
re
tri
e
v
a
l
sy
ste
m
,
”
In
t.
J
.
Co
mp
u
t.
S
c
i.
I
n
f.
T
e
c
h
n
o
l.
,
v
o
l.
7
,
n
o
.
1
,
p
p
.
4
0
7
-
4
1
2
,
2
0
1
6
.
[1
9
]
J.
Hu
a
n
g
,
e
t
a
l.
,
“
I
m
a
g
e
in
d
e
x
in
g
u
sin
g
c
o
lo
r
c
o
rre
lo
g
ra
m
s,”
Pro
c
e
e
d
in
g
s
o
f
IEE
E
c
o
mp
u
ter
so
c
iety
c
o
n
fer
e
n
c
e
o
n
Co
mp
u
ter
V
isio
n
a
n
d
Pa
t
ter
n
Rec
o
g
n
it
io
n
,
p
p
.
7
6
2
-
7
6
8
,
1
9
9
7
.
[2
0
]
D.
G
.
L
o
w
e
,
“
Distin
c
ti
v
e
I
m
a
g
e
F
e
a
tu
re
s
f
ro
m
S
c
a
le
-
In
v
a
r
ian
t
Ke
y
p
o
in
ts
A
b
stra
c
t
b
y
M
a
tt
h
ij
s
Do
rst
,
”
In
ter
n
a
t
io
n
a
l
jo
u
rn
a
l
o
f
c
o
mp
u
ter
v
isio
n
,
v
o
l.
6
0
,
n
o
.
2
,
p
p
.
9
1
-
1
1
0
,
2
0
0
4
[2
1
]
J.
Zh
a
n
g
,
e
t
a
l.
,
“
L
o
c
a
l
f
e
a
tu
re
s
a
n
d
k
e
rn
e
ls
f
o
r
c
la
ss
i
f
ica
ti
o
n
o
f
tex
tu
re
a
n
d
o
b
jec
t
c
a
teg
o
ries
:
A
c
o
m
p
re
h
e
n
siv
e
stu
d
y
,
”
In
ter
n
a
ti
o
n
a
l
j
o
u
rn
a
l
o
f
c
o
mp
u
ter
v
isi
o
n
,
v
o
l.
7
3
,
n
o
.
2
,
p
p
.
2
1
3
-
2
3
8
,
2
0
0
7
.
[2
2
]
B.
S
.
M
a
n
ju
n
a
th
,
“
T
e
x
tu
re
f
e
a
tu
r
e
s
f
o
r
b
ro
w
sin
g
a
n
d
re
tri
e
v
a
l
o
f
im
a
g
e
d
a
ta,”
IEE
E
T
ra
n
s.
P
a
tt
e
r
n
An
a
l.
M
a
c
h
.
In
tell.
,
v
o
l.
1
8
,
n
o
.
8
,
p
p
.
8
3
7
-
8
4
2
,
1
9
9
6
.
[2
3
]
X
.
Ch
e
n
,
J.
He
,
R.
L
a
w
re
n
c
e
,
a
n
d
J.
G
.
C
a
rb
o
n
e
ll
,
“
A
d
a
p
ti
v
e
m
u
lt
i
-
tas
k
sp
a
rse
lea
rn
in
g
w
it
h
a
n
a
p
p
li
c
a
ti
o
n
t
o
f
M
RI
stu
d
y
,
”
Pro
c
.
1
2
th
S
IAM
In
t
.
Co
n
f.
Da
t
a
M
in
i
n
g
,
S
DM
2
0
1
2
,
p
p
.
2
1
2
-
2
2
3
,
2
0
1
2
.
[2
4
]
Y.
Zh
a
n
g
a
n
d
Q.
Ya
n
g
,
"
A
S
u
rv
e
y
o
n
M
u
lt
i
-
T
a
sk
L
e
a
rn
in
g
,
"
a
rXiv p
re
p
rin
t
a
rXiv:
1
7
0
7
.
0
8
1
1
4
,
2
0
1
7
.
[2
5
]
J.
L
e
e
Ro
d
g
e
rs
a
n
d
W
.
A
lan
Nic
e
W
a
n
d
e
r,
“
T
h
irt
e
e
n
w
a
y
s
to
lo
o
k
a
t
th
e
c
o
rre
l
a
ti
o
n
c
o
e
ff
icie
n
t,
”
T
h
e
Ame
ric
a
n
S
ta
ti
st
icia
n
,
v
o
l.
4
2
,
n
o
.
1
,
p
p
.
5
9
-
6
6
,
1
9
8
8
.
[2
6
]
T.
-
S
.
Ch
u
a
,
e
t
a
l.
,
"
NUS
-
W
IDE:
A
Re
a
l
-
W
o
rld
Web
I
m
a
g
e
Da
t
a
b
a
se
f
ro
m
N
a
ti
o
n
a
l
Un
iv
e
rsit
y
o
f
S
in
g
a
p
o
re
,
"
Pro
c
e
e
d
in
g
s
o
f
t
h
e
ACM
in
ter
n
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
ima
g
e
a
n
d
v
id
e
o
re
trie
v
a
l,
2
0
0
9
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
P
iy
a
w
a
t
S
a
e
n
g
p
e
tc
h
,
h
e
re
c
e
iv
e
d
B
.
S
.
d
e
g
re
e
f
ro
m
th
e
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
t
S
o
n
g
k
h
la
Ra
jab
h
a
t
U
n
iv
e
rsity
,
T
h
a
il
a
n
d
,
in
1
9
9
4
,
M
S
c
f
ro
m
th
e
P
rin
c
e
o
f
S
o
n
g
k
h
la
Un
iv
e
rsit
y
,
T
h
a
il
a
n
d
i
n
2
0
0
5
,
a
n
d
th
e
P
h
.
D
.
d
e
g
r
e
e
in
2
0
1
4
f
ro
m
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
a
n
d
In
f
o
rm
a
ti
o
n
S
c
ien
c
e
a
t
Ki
n
g
M
o
n
g
k
u
t'
s
Un
iv
e
rsit
y
o
f
Tec
h
n
o
lo
g
y
No
rth
Ba
n
g
k
o
k
,
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
.
S
in
c
e
2
0
1
0
,
h
e
h
a
s
w
o
rk
e
d
in
t
h
e
Co
m
p
u
ter
S
c
ien
c
e
P
r
o
g
ra
m
a
t
S
u
ra
tt
h
a
n
i
Ra
jab
h
a
t
Un
iv
e
rsit
y
,
T
h
a
il
a
n
d
.
His c
u
rre
n
t
re
se
a
rc
h
in
M
L
a
n
d
A
I.
Lu
e
p
o
l
Pi
p
a
n
m
a
e
k
a
p
o
r
n
is
is
c
u
rre
n
tl
y
a
lec
tu
re
r
a
t
th
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
a
n
d
In
f
o
rm
a
ti
o
n
S
c
ien
c
e
,
Ki
n
g
M
o
n
g
k
u
t'
s
Un
iv
e
rsit
y
o
f
No
rth
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
.
He
h
o
l
d
s
b
a
c
h
e
lo
r'
s
a
n
d
m
a
ste
r'
s
d
e
g
re
e
s
in
c
o
m
p
u
ter
sc
ien
c
e
.
He
a
lso
e
a
rn
e
d
h
is
d
o
c
t
o
ra
l
d
e
g
re
e
i
n
c
o
m
p
u
ter
sc
ien
c
e
f
ro
m
Qu
e
e
n
sla
n
d
Un
iv
e
rsity
o
f
T
e
c
h
n
o
lo
g
y
,
Au
stra
li
a
,
g
ra
d
u
a
ti
n
g
in
2
0
1
3
.
His c
u
rre
n
t
re
se
a
rc
h
in
tere
sts in
v
o
lv
e
in
f
o
rm
a
ti
o
n
re
tri
e
v
a
l,
we
b
m
in
in
g
,
a
n
d
d
a
ta m
in
in
g
.
S
u
w
a
tc
h
a
i
K
a
m
o
l
sa
n
tiro
j
is
c
u
rre
n
tl
y
a
l
e
c
tu
re
r
a
t
th
e
D
e
p
a
rtme
n
t
o
f
Co
m
p
u
ter
a
n
d
In
f
o
rm
a
ti
o
n
S
c
ien
c
e
,
Ki
n
g
M
o
n
g
k
u
t'
s
Un
iv
e
rsit
y
o
f
No
rth
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
.
He
h
o
l
d
s
m
a
ste
r'
s
d
e
g
re
e
s
in
c
o
m
p
u
ter
sc
ien
c
e
.
He
a
lso
e
a
rn
e
d
h
is
d
o
c
to
ra
l
d
e
g
re
e
in
c
o
m
p
u
ter
e
n
g
in
e
e
rin
g
f
ro
m
Ka
s
e
tsa
rt
Un
iv
e
rsit
y
,
T
h
a
i
lan
d
,
g
ra
d
u
a
ti
n
g
in
2
0
1
3
.
His
c
u
rre
n
t
re
se
a
rc
h
in
tere
sts in
v
o
lv
e
f
u
z
z
y
lo
g
ic,
n
e
u
ra
l
n
e
tw
o
rk
,
a
n
d
a
rti
f
icia
l
in
telli
g
e
n
c
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.