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.
9
,
No
.
1
,
Ap
r
il
2
0
2
0
,
p
p
.
4
6
~
5
6
I
SS
N:
2252
-
8
7
7
6
,
DOI
: 1
0
.
1
1
5
9
1
/iji
ct.
v
9
i
1
.
p
p
4
6
-
5
6
46
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
No
v
a
l adv
a
nce non
-
linea
r d
escripto
r and cha
ra
cte
rist
ic
equiv
a
lency
t
o
pr
edict
CT from
MRI
ima
g
e
So
wm
y
a
B
a
chu,
Sh
rut
ibh
a
r
g
a
v
a
Cho
ub
ey
,
Abhi
s
hek
Cho
ub
ey
S
re
e
n
id
h
i
In
sti
tu
te
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
In
d
ia
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
28
,
2
0
19
R
ev
is
ed
J
an
12
,
2
0
20
Acc
ep
ted
Feb
11
,
2
0
20
Atten
u
a
ti
o
n
c
o
rre
c
ti
o
n
d
e
sig
n
e
d
fo
r
P
ET
/
M
R
h
y
b
r
id
ima
g
in
g
f
ra
m
e
wo
rk
s
a
lo
n
g
with
p
o
r
ti
o
n
m
a
k
i
n
g
a
rr
a
n
g
e
m
e
n
ts
u
se
d
f
o
r
M
R
-
b
a
se
d
ra
d
iatio
n
trea
tme
n
t
re
m
a
in
tes
ti
n
g
b
e
c
a
u
se
o
f
lac
k
in
g
h
i
g
h
-
e
n
e
rg
y
p
h
o
to
n
we
a
k
e
n
in
g
d
a
ta.
We
p
re
se
n
t
a
n
e
w
m
e
th
o
d
s
o
a
s
to
u
se
s
t
h
e
lea
rn
e
d
n
o
n
li
n
e
a
r
n
e
ig
h
b
o
r
h
o
o
d
d
e
sc
rip
t
o
rs
a
lso
h
i
g
h
li
g
h
t
c
o
o
r
d
in
a
t
in
g
to
wa
rd
fo
re
s
e
e
p
se
u
d
o
-
CT
p
ict
u
re
s
sta
rti
n
g
T1
w
a
lo
n
g
with
T
2
w
M
RI
i
n
fo
rm
a
ti
o
n
.
T
h
e
n
o
n
li
n
e
a
r
n
e
ig
h
b
o
r
h
o
o
d
d
e
sc
rip
to
rs
a
re
a
c
q
u
ired
th
r
o
u
g
h
a
n
t
icip
a
ti
n
g
th
e
d
irec
t
d
e
sc
rip
to
rs
in
tere
ste
d
in
th
e
n
o
n
li
n
e
a
r
h
ig
h
-
d
ime
n
si
o
n
a
l
sp
a
c
e
u
t
il
izin
g
a
n
u
n
e
q
u
iv
o
c
a
l
c
o
n
stit
u
e
n
t
g
u
i
d
e
a
lso
l
o
w
-
p
o
si
ti
o
n
g
u
e
ss
t
h
ro
u
g
h
re
g
u
late
d
c
o
m
p
lex
re
g
u
lariz
a
ti
o
n
.
Th
e
n
e
a
rb
y
n
e
ig
h
b
o
rs
o
f
e
v
e
r
y
n
e
a
r
d
e
sc
rip
to
r
in
si
d
e
th
e
d
a
ta
M
R
p
ictu
re
s
a
re
lo
o
k
e
d
d
u
ri
n
g
a
n
o
b
li
g
e
d
sp
a
ti
a
l
e
x
te
n
t
o
f
th
e
M
R
p
ictu
re
s
a
m
o
n
g
th
e
train
i
n
g
d
a
tas
e
t.
By
t
h
a
t
p
o
i
n
t,
th
e
p
se
u
d
o
-
CT
p
a
tch
e
s
a
re
e
v
a
lu
a
ted
th
r
o
u
g
h
k
-
c
lo
se
st
n
e
ig
h
b
o
r
re
lap
se
.
Th
e
p
l
a
n
n
e
d
p
ro
c
e
d
u
re
d
e
sig
n
e
d
f
o
r
p
se
u
d
o
-
CT
f
o
re
c
a
st
is
q
u
a
n
ti
tati
v
e
ly
b
r
o
k
e
d
o
wn
wa
rd
o
n
t
o
p
o
f
a
d
a
tas
e
t
c
o
m
p
risi
n
g
o
f
c
o
o
r
d
in
a
t
e
d
m
in
d
M
RI
a
lo
n
g
with
CT
p
ict
u
re
s
o
n
o
r
a
fter 1
3
su
b
jec
ts
.
K
ey
w
o
r
d
s
:
C
T
ca
lcu
latio
n
KNN
r
eg
r
ess
io
n
L
o
w
-
r
an
k
ap
p
r
o
x
im
atio
n
No
n
-
lin
ea
r
d
escr
ip
to
r
PET
atten
u
atio
n
co
r
r
ec
tio
n
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
:
So
wm
y
a
B
ac
h
u
,
Sre
en
id
h
i
I
n
s
titu
te
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
,
Yam
n
am
p
et,
Gh
atk
esar
,
Hy
d
e
r
ab
ad
,
T
ela
n
g
an
a
5
0
1
3
0
1
,
I
n
d
ia
.
E
m
ail:
Sb
3
0
3
1
0
7
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Nu
m
er
o
u
s
in
n
o
v
ativ
e
m
eth
o
d
s
co
n
tain
b
e
p
la
n
n
ed
in
f
av
o
r
o
f
p
r
ed
ictin
g
C
T
im
ag
e
r
y
s
tar
tin
g
MRI
d
ata
,
as
well
as
co
n
tain
er
,
b
e
class
if
y
in
ter
ested
in
f
o
u
r
class
es
:
s
eg
m
en
tatio
n
,
atlas,
ex
ac
t
s
er
ies
-
also
p
atch
-
b
ased
tech
n
iq
u
e
s
.
I
n
s
id
e
t
h
e
s
eg
m
en
tatio
n
-
b
ased
tech
n
i
q
u
e
s
[
1
-
3
]
,
m
a
g
n
etic
r
eso
n
an
ce
im
ag
er
y
is
s
eg
m
en
te
d
in
ter
ested
in
s
id
e
d
is
s
im
i
lar
h
an
k
ie
less
o
n
s
(
e.
g
.
,
s
o
f
t
tis
s
u
e,
f
at,
an
d
atm
o
s
p
h
er
e,
alo
n
g
with
clea
n
)
.
E
v
er
y
g
r
o
u
p
is
n
ex
t
ass
ig
n
ed
t
h
e
p
r
e
-
d
ef
i
n
ed
r
e
d
u
ctio
n
co
ef
f
icien
ts
(
L
AC
)
o
th
er
wis
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ass
es
s
m
en
t
s
.
T
o
war
d
g
et
co
r
r
ec
t
s
eg
m
en
tatio
n
,
th
e
f
u
z
zy
cl
u
s
ter
in
g
m
eth
o
d
alo
n
g
with
S
PM8
s
o
f
twar
e
we
r
e
u
tili
ze
d
in
s
id
e
[
1
,
2
]
also
[
3
]
,
co
r
r
esp
o
n
d
in
g
ly
.
T
h
e
ac
cu
r
ate
n
ess
o
f
th
e
s
eg
m
en
tatio
n
-
b
ased
m
eth
o
d
s
d
esig
n
ed
f
o
r
p
C
T
ca
lcu
latio
n
b
e
in
co
m
p
lete
s
in
ce
th
e
s
eg
m
en
ted
tis
s
u
e
r
eg
io
n
s
co
n
tr
ib
u
te
in
t
h
e
d
ir
ec
tio
n
o
f
th
e
s
im
ilar
p
r
e
-
d
ef
i
n
ed
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ass
ess
m
en
t
also
th
e
v
a
r
iatio
n
in
s
id
e
th
e
p
r
o
p
e
r
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
ass
es
s
m
en
t
s
u
r
r
o
u
n
d
ed
th
r
o
u
g
h
th
e
s
im
ilar
h
an
k
ie
b
e
u
n
s
ee
n
.
B
asic
f
u
n
d
am
en
tal
d
esig
n
o
f
t
h
e
atlas
-
b
ased
tech
n
iq
u
e
s
[
4
-
6
]
ex
is
ts
s
tr
aig
h
tf
o
r
war
d
.
A
d
at
aset
s
o
as
to
co
n
tain
n
u
m
er
o
u
s
m
ag
n
e
ti
c
r
eso
n
an
ce
/
co
m
p
u
ted
to
m
o
g
r
ap
h
y
im
ag
e
p
air
s
b
e
n
ec
ess
ar
y
.
I
n
itial,
an
atlas
d
ataset
b
e
r
e
g
is
ter
to
war
d
an
in
p
u
t
m
a
g
n
etic
r
eso
n
an
ce
im
a
g
in
g
p
ictu
r
e
th
r
o
u
g
h
ca
lcu
lat
i
n
g
th
e
d
e
f
o
r
m
atio
n
f
ield
am
o
n
g
th
e
atlas
alo
n
g
with
th
e
m
a
g
n
etic
r
eso
n
an
ce
im
ag
in
g
p
ictu
r
e.
N
ex
t,
t
h
e
e
q
u
iv
alen
t
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
im
ag
er
y
b
e
d
is
to
r
ted
to
war
d
th
is
MRI
im
ag
e
b
y
m
ea
n
s
o
f
th
e
m
u
lti
-
atlas
in
f
o
r
m
atio
n
d
is
s
em
in
atio
n
p
lan
i
n
tr
o
d
u
ce
d
with
in
[
5
,
6
]
.
F
in
ally
,
t
h
e
o
b
t
ain
ed
C
T
im
ag
er
y
is
m
er
g
ed
i
n
ter
ested
in
th
e
C
T
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
N
o
va
l a
d
va
n
ce
n
o
n
-
lin
ea
r
d
es
crip
to
r
a
n
d
ch
a
r
a
cteris
tic
eq
u
iva
len
cy
to
…
(
S
o
w
mya
B
a
ch
u
)
47
ca
lcu
latio
n
.
W
ith
in
th
e
p
ictu
r
e
s
y
n
th
esis
s
tag
e
,
th
e
Gau
s
s
ian
p
r
o
ce
d
u
r
e
wea
k
en
i
n
g
[
5
]
,
lim
ited
p
ictu
r
e
co
m
p
ar
is
o
n
-
b
ased
ad
v
an
ce
[
5
]
also
v
o
x
el
-
wis
e
m
ax
im
u
m
lik
elih
o
o
d
s
tr
en
g
th
av
e
r
ag
in
g
ad
v
an
ce
s
[
6
]
b
e
u
tili
ze
d
also
v
alid
ate.
T
h
e
p
r
esen
tatio
n
o
f
th
e
atlas
-
b
ased
m
eth
o
d
s
d
ep
en
d
s
p
o
wer
f
u
ll
y
u
n
d
er
t
o
p
o
f
t
h
e
r
eg
is
tr
atio
n
ex
ac
tn
ess
alo
n
g
with
th
e
en
d
u
r
in
g
p
o
p
u
latio
n
s
en
co
m
p
ass
th
r
o
u
g
h
th
e
atlas.
H
o
wev
er
,
th
e
atlas
m
ig
h
t f
ail
to
co
r
r
esp
o
n
d
to
war
d
th
e
r
ed
u
ctio
n
o
f
p
atien
ts
wh
o
co
n
tain
a
p
o
r
tio
n
o
f
th
eir
s
k
u
lls
d
etac
h
ed
.
S
u
itab
le
to
w
ar
d
th
e
p
ar
ticu
lar
l
y
to
wer
in
g
r
ed
u
ctio
n
ass
ess
m
en
t
o
f
th
e
s
k
eleto
n
,
th
e
ex
ac
t
ca
lcu
latio
n
o
f
clea
n
in
s
id
e
p
r
ed
icted
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
is
p
r
ef
er
r
ed
.
T
h
is
r
e
q
u
ir
e
to
f
ac
ilit
ate
th
e
m
ag
n
etic
r
eso
n
an
ce
p
ictu
r
e
co
n
tain
er
p
r
o
v
id
e
s
u
f
f
icien
t
clu
e
t
o
war
d
r
ec
o
g
n
i
ze
th
e
s
k
eleto
n
.
T
h
o
u
g
h
,
p
r
ed
ictab
le
m
ag
n
etic
r
eso
n
an
ce
im
ag
i
n
g
s
ig
n
al
b
e
ess
en
tially
in
ap
p
r
o
p
r
iate
in
f
av
o
r
o
f
d
ep
ictin
g
b
o
n
es
as
well
as
atm
o
s
p
h
er
e
s
tr
u
ctu
r
es
d
u
e
to
th
e
p
r
o
to
n
d
en
s
ity
d
o
m
in
an
ce
also
th
e
m
ag
n
etic
r
eso
n
an
ce
im
a
g
in
g
s
ig
n
al
r
ec
r
ea
tio
n
m
ec
h
an
is
m
.
I
n
r
ec
en
t
tim
es,
n
u
m
er
o
u
s
m
eth
o
d
s
b
ased
on
to
p
o
f
ex
ac
t
im
a
g
in
g
s
eq
u
e
n
ce
s
in
clu
d
e
ar
e
p
la
n
n
ed
[7
-
1
2
]
.
Ultr
a
s
h
o
r
t
ec
h
o
tim
e
s
eq
u
en
ce
(
UT
E
)
also
ze
r
o
ec
h
o
tim
e
s
eq
u
en
ce
(
Z
T
E
)
was
u
s
e
to
war
d
im
p
r
o
v
es
th
e
b
o
n
e
r
ec
o
g
n
itio
n
in
s
id
e
[7
-
1
1
]
alo
n
g
with
[
1
2
]
,
co
r
r
esp
o
n
d
in
g
l
y
.
T
h
e
m
ajo
r
c
o
n
s
tr
ain
t
o
f
th
is
m
eth
o
d
is
th
e
f
u
r
th
e
r
in
s
tan
ce
r
ate
with
in
p
r
o
g
r
ess
io
n
in
f
o
r
m
atio
n
attai
n
m
en
t in
s
id
e
th
e
m
e
d
ical
ap
p
li
ca
tio
n
s
.
D
u
r
in
g
th
is
lear
n
in
g
,
a
p
atch
-
b
ased
tech
n
iq
u
e
ca
lled
ch
a
r
a
cter
is
tic
co
r
r
es
p
o
n
d
in
g
th
r
o
u
g
h
lear
n
e
d
non
-
lin
ea
r
d
escr
ip
to
r
s
(
FML
ND)
is
p
lan
n
ed
o
n
b
eh
alf
o
f
p
r
ed
ictin
g
p
C
T
s
tar
tin
g
MRI
r
ec
o
r
d
s
.
T
o
war
d
g
et
b
etter
th
e
ab
ilit
y
o
f
t
h
e
clea
n
r
ec
o
g
n
itio
n
,
a
m
i
x
tu
r
e
o
f
th
ick
s
ca
le
-
in
v
ar
ian
t
f
ea
tu
r
e
tr
an
s
f
o
r
m
s
(
SIFT
)
d
escr
ip
to
r
s
b
y
n
o
r
m
aliz
e
u
n
p
r
o
ce
s
s
ed
ar
ea
s
ar
e
u
s
ed
b
ec
a
u
s
e
th
e
m
ain
d
escr
ip
to
r
o
f
m
ag
n
etic
r
eso
n
an
ce
im
ag
er
y
r
elativ
ely
th
a
n
m
ag
n
etic
r
eso
n
an
ce
r
aw
p
atch
es
o
th
er
wis
e
v
o
x
els
s
in
ce
h
er
e
[
4
,
5
,
8
,
1
0
,
1
3
,
1
4
,
1
6
]
.
Scale
I
n
v
ar
ian
t
Featu
r
e
T
r
an
s
f
o
r
m
ch
ar
ac
ter
is
tic
d
ep
icts
s
tr
u
ctu
r
al
in
f
o
r
m
atio
n
,
wh
ich
is
p
r
ec
io
u
s
in
s
id
e
r
ec
o
g
n
izi
n
g
clea
n
h
a
n
k
ie
alo
n
g
with
atm
o
s
p
h
er
e
in
s
id
e
m
a
g
n
etic
r
eso
n
an
ce
im
ag
in
g
in
f
o
r
m
atio
n
.
O
n
th
e
way
to
b
etter
h
o
ld
th
e
n
o
n
-
lin
e
ar
ity
o
f
m
ap
p
in
g
am
o
n
g
th
e
m
ain
d
escr
ip
to
r
s
alo
n
g
with
th
e
c
o
m
p
u
ted
t
o
m
o
g
r
ap
h
y
u
n
p
r
o
c
ess
ed
ar
ea
s
,
th
e
m
ain
d
escr
ip
to
r
s
ar
e
p
r
ed
ictab
le
to
war
d
a
h
ig
h
-
d
im
e
n
s
io
n
al
g
ap
b
y
m
ea
n
s
o
f
o
p
en
f
ea
tu
r
e
m
ap
s
[
1
7
]
to
war
d
g
et
wid
esp
r
ea
d
MRI
in
f
o
r
m
atio
n
.
W
h
en
r
ec
o
m
m
en
d
ed
with
in
o
u
r
p
r
ec
e
d
in
g
lear
n
[
1
8
]
,
th
e
m
ap
p
in
g
s
tar
tin
g
t
h
e
m
ain
d
escr
ip
to
r
s
t
o
war
d
th
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
u
n
p
r
o
ce
s
s
ed
p
atch
es
co
n
tain
er
e
x
is
t
r
o
u
g
h
ly
m
ea
s
u
r
ed
b
ec
au
s
e
n
ea
r
b
y
li
n
ea
r
b
elo
w
th
e
n
ea
r
b
y
s
p
atial
lim
it
atio
n
.
A
s
u
p
e
r
v
is
ed
lear
n
in
g
tech
n
iq
u
e
is
p
la
n
n
ed
o
n
b
e
h
alf
o
f
en
s
u
r
e
th
e
v
iab
ilit
y
o
f
th
e
o
v
er
s
tatem
en
t
th
r
o
u
g
h
lear
n
a
lo
ca
l
non
lin
ea
r
d
escr
ip
to
r
(
L
ND)
,
wh
ich
b
e
a
wid
esp
r
ea
d
l
o
w
-
r
a
n
k
esti
m
ate
o
f
n
o
n
-
lin
ea
r
d
escr
ip
to
r
s
.
D
u
r
in
g
th
is
k
n
o
wled
g
e
c
o
n
s
tr
u
ctio
n
,
t
h
e
c
o
m
p
ar
is
o
n
in
f
o
r
m
atio
n
o
f
th
e
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
ar
ea
s
is
u
s
ed
o
n
b
eh
alf
o
f
r
eg
u
lar
izatio
n
al
o
n
g
with
s
u
p
e
r
v
is
es th
e
d
im
en
s
io
n
ality
d
ec
r
ea
s
e
o
f
th
e
lear
n
e
d
n
o
n
-
lin
ea
r
d
escr
ip
to
r
s
.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
C
alcu
latio
n
s
o
f
C
T
s
u
b
s
titu
te
s
s
tar
tin
g
MR
im
ag
er
y
ar
e
cl
in
ically
p
r
ef
er
r
ed
u
s
ed
i
n
f
av
o
r
o
f
d
o
s
e
p
r
ep
ar
atio
n
in
s
id
e
MR
-
b
ased
r
a
d
iatio
n
t
h
er
ap
y
alo
n
g
with
r
ed
u
ctio
n
alter
atio
n
wi
th
in
PET
/MR
[
1
3
]
.
A
llo
win
g
f
o
r
t
h
at
p
r
esen
t
is
n
o
wo
r
ld
wid
e
r
elatio
n
am
o
n
g
in
ten
s
ities
with
in
MR
alo
n
g
with
C
T
im
ag
er
y
,
w
e
s
u
g
g
est
lo
ca
l
s
p
ar
s
e
co
r
r
esp
o
n
d
en
ce
co
m
b
i
n
atio
n
(
L
SC
C
)
u
s
ed
in
s
u
p
p
o
r
t
o
f
th
e
ca
lcu
latio
n
o
f
C
T
s
u
b
s
titu
te
o
n
o
r
af
ter
MR
im
ag
er
y
.
D
u
r
i
n
g
L
SC
C
,
we
s
u
p
p
o
s
e
to
MR
as
well
a
s
C
T
p
atch
es
ar
e
s
it
u
ated
u
n
d
e
r
to
p
o
f
two
n
o
n
lin
ea
r
m
an
i
f
o
ld
s
alo
n
g
with
th
e
m
ap
p
in
g
o
n
o
r
af
ter
th
e
m
ag
n
etic
r
eso
n
an
ce
d
iv
er
s
e
to
war
d
th
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
d
iv
er
s
e
ap
p
r
o
x
im
ates
a
d
if
f
eo
m
o
r
p
h
i
s
m
in
a
r
estricte
d
r
est
r
ain
t.
A
n
u
m
b
er
o
f
m
eth
o
d
s
ar
e
u
s
ed
to
war
d
co
n
s
tr
ict
r
e
g
i
o
n
:
1
)
d
esig
n
ed
o
n
b
eh
alf
o
f
ev
er
y
a
r
ea
in
s
id
e
th
e
test
MR
p
ictu
r
e,
a
lim
ited
in
v
esti
g
ate
g
ap
is
u
s
ed
t
o
war
d
tak
e
o
u
t
p
atch
es
as
o
f
th
e
p
r
e
p
ar
atio
n
MR/C
T
p
air
s
to
war
d
b
u
ild
MR
as
well
as
co
m
p
u
te
d
to
m
o
g
r
a
p
h
y
d
i
ctio
n
ar
y
;
2
)
k
-
Nea
r
est
Neig
h
b
o
r
s
is
u
tili
ze
d
to
war
d
lim
it
r
eg
io
n
in
s
id
e
th
e
m
ag
n
etic
r
eso
n
an
ce
v
o
ca
b
u
lar
y
;
3
)
o
u
tlier
r
ec
o
g
n
itio
n
is
p
er
f
o
r
m
ed
to
war
d
lim
it
r
eg
io
n
in
s
id
e
th
e
c
o
m
p
u
te
d
to
m
o
g
r
a
p
h
y
d
ictio
n
ar
y
;
4
)
li
m
ited
f
asten
E
m
b
ed
d
in
g
is
u
s
ed
to
wa
r
d
r
eso
lv
in
g
th
e
m
ag
n
etic
r
eso
n
an
ce
d
ictio
n
ar
y
c
o
ef
f
icien
ts
af
ter
r
ep
r
esen
tin
g
th
e
m
ag
n
etic
r
eso
n
an
ce
d
if
f
icu
lt
ex
a
m
p
le.
B
el
o
w
th
ese
r
estricte
d
co
n
s
tr
ain
ts
,
th
e
co
ef
f
icien
t
weig
h
ts
ar
e
lin
ea
r
ly
tr
an
s
f
er
r
ed
o
n
o
r
af
ter
m
ag
n
etic
r
eso
n
an
c
e
to
war
d
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
also
u
s
ed
to
jo
in
th
e
s
am
p
les
with
in
th
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
d
ictio
n
ar
y
to
war
d
m
a
k
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ca
lcu
la
tio
n
s
.
T
h
e
p
lan
n
ed
tech
n
iq
u
es
h
av
e
b
ee
n
ev
alu
ate
d
o
n
b
eh
al
f
o
f
m
i
n
d
im
ag
e
r
y
u
n
d
er
t
o
p
o
f
a
d
ataset
o
f
1
3
s
u
b
jects.
E
v
er
y
t
o
p
ic
h
as
T
1
-
as
well
as
T
2
-
weig
h
ted
m
ag
n
etic
r
eso
n
a
n
ce
im
ag
er
y
,
also
a
c
o
m
p
u
ted
to
m
o
g
r
ap
h
y
p
ictu
r
e
th
r
o
u
g
h
a
s
u
m
o
f
3
9
im
ag
er
y
.
R
ed
u
ctio
n
alter
atio
n
is
s
ig
n
if
ican
t
d
esig
n
ed
in
f
av
o
r
o
f
p
o
s
itro
n
em
is
s
io
n
to
m
o
g
r
ap
h
y
r
eb
u
ild
in
g
[
1
4
]
.
D
u
r
i
n
g
PET
/MR,
m
ag
n
etic
r
eso
n
an
ce
in
ten
s
ities
ar
e
n
o
t
s
tr
aig
h
t
co
r
r
elate
d
to
war
d
r
ed
u
ctio
n
co
ef
f
icien
ts
th
at
ar
e
d
esire
d
i
n
s
id
e
p
o
s
itro
n
em
is
s
io
n
to
m
o
g
r
ap
h
y
im
ag
in
g
.
T
h
e
r
ed
u
cti
o
n
co
ef
f
icien
t
p
lan
co
n
tain
er
b
e
r
esu
ltin
g
in
c
o
m
p
u
ted
to
m
o
g
r
a
p
h
y
im
ag
e
r
y
.
C
o
n
s
eq
u
en
tly
,
ca
lcu
latio
n
o
f
co
m
p
u
ted
to
m
o
g
r
ap
h
y
s
u
b
s
titu
te
as
o
f
m
a
g
n
etic
r
eso
n
an
ce
im
ag
er
y
is
p
r
ef
er
r
e
d
o
n
b
eh
alf
o
f
r
ed
u
ctio
n
alter
at
io
n
in
s
id
e
p
o
s
itro
n
em
is
s
io
n
to
m
o
g
r
a
p
h
y
/m
a
g
n
eti
c
r
eso
n
an
ce
.
Me
th
o
d
s
:
T
h
is
le
ar
n
in
g
p
r
esen
ts
a
p
atch
-
b
ased
tech
n
iq
u
e
u
s
ed
o
n
b
eh
alf
o
f
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
ca
lcu
latio
n
b
eg
in
n
i
n
g
MR
im
ag
er
y
; g
e
n
er
ate
r
e
d
u
ctio
n
m
ap
s
u
s
ed
in
f
av
o
r
o
f
p
o
s
itro
n
em
is
s
io
n
to
m
o
g
r
ap
h
y
r
eb
u
ild
i
n
g
.
S
in
ce
n
o
wo
r
l
d
wid
e
r
elatio
n
ex
is
ts
am
o
n
g
m
ag
n
etic
r
eso
n
an
ce
as
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.
9
,
No
.
1
,
Ap
r
il
20
20
:
4
6
–
5
6
48
well
as
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
in
ten
s
ities
,
we
s
u
g
g
est
lim
ited
d
if
f
e
o
m
o
r
p
h
ic
m
a
p
p
in
g
(
L
D
M)
u
s
ed
in
f
av
o
r
o
f
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ca
lc
u
latio
n
.
D
u
r
in
g
L
DM
,
we
s
u
p
p
o
s
e
s
o
as
to
MR
as
well
as
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
p
atch
es
ar
e
s
itu
ated
u
n
d
er
to
p
o
f
two
n
o
n
-
li
n
ea
r
m
an
if
o
l
d
s
alo
n
g
with
th
e
m
ap
p
in
g
as
o
f
th
e
m
ag
n
etic
r
eso
n
an
ce
v
ar
io
u
s
to
war
d
th
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
v
ar
io
u
s
ap
p
r
o
x
im
ates
a
d
if
f
e
o
m
o
r
p
h
is
m
in
a
lim
ited
r
estrictio
n
.
R
eg
io
n
is
s
ig
n
if
ican
t w
ith
in
L
DM
alo
n
g
with
is
c
o
n
tr
o
lled
with
m
ea
n
s
o
f
th
e
n
e
x
t te
ch
n
iq
u
es.
I
n
ad
d
itio
n
,
d
u
r
i
n
g
th
e
n
o
v
el
PET
as
well
as
m
ag
n
etic
r
eso
n
an
ce
im
ag
i
n
g
s
ca
n
n
er
,
s
im
p
l
y
m
ag
n
etic
r
eso
n
an
ce
im
a
g
er
y
ar
e
o
b
tain
ab
le,
wh
ich
ar
e
alas
n
o
t
o
p
en
ly
ap
p
r
o
p
r
iate
to
war
d
r
ed
u
ctio
n
alter
atio
n
.
T
h
ese
is
s
u
es
r
ea
lly
p
r
o
m
p
t
th
e
ex
p
an
s
io
n
o
f
m
et
h
o
d
s
u
s
ed
o
n
b
eh
alf
o
f
d
ep
en
d
ab
le
esti
m
atio
n
o
f
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
[
1
9
]
p
ictu
r
e
o
n
o
r
af
ter
its
eq
u
i
v
alen
t
m
a
g
n
etic
r
eso
n
an
ce
p
ictu
r
e
o
f
a
s
im
ila
r
to
p
ic.
D
u
r
i
n
g
th
is
ar
ticle,
we
s
u
g
g
est
a
lear
n
i
n
g
-
b
ased
tech
n
iq
u
e
to
u
n
d
e
r
tak
e
t
h
is
d
em
an
d
in
g
d
if
f
icu
lty
.
I
n
p
ar
ticu
lar
,
we
in
itial
s
ep
ar
atio
n
a
k
n
o
wn
m
ag
n
etic
r
eso
n
an
ce
p
ictu
r
e
in
ter
ested
i
n
a
lo
ca
te
o
f
p
atch
es.
A
f
ter
th
at,
m
ea
n
t
o
n
b
eh
alf
o
f
ev
er
y
p
atch
,
we
u
tili
ze
th
e
p
r
ep
ar
e
d
ac
cid
en
tal
ju
n
g
le
to
war
d
s
tr
aig
h
t
f
o
r
ec
ast
a
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
p
atch
s
in
ce
a
p
r
ep
ar
ed
o
u
t
p
u
t,
an
y
wh
er
e
a
n
o
v
el
co
llectio
n
m
o
d
el
is
as we
l
l u
s
ed
to
war
d
m
ak
e
s
u
r
e
th
e
s
tr
o
n
g
ca
lcu
latio
n
.
An
item
d
etec
tio
n
s
ch
e
m
e
h
a
s
be
en
d
ev
elo
p
ed
to
u
tili
ze
a
n
o
v
el
g
r
o
u
p
o
f
lim
ited
p
ictu
r
e
f
ea
tu
r
es
[
1
6
]
.
T
h
e
f
ea
tu
r
es
ar
e
i
n
v
ar
ia
n
t
to
war
d
p
ictu
r
e
s
ca
lin
g
,
co
n
v
er
s
io
n
,
alo
n
g
with
r
o
tatio
n
,
also
in
co
m
p
letely
in
v
ar
ian
t
to
war
d
lig
h
tin
g
ch
an
g
e
also
af
f
in
e
o
th
e
r
wis
e
3
D
r
id
g
e
T
h
ese
f
ea
tu
r
es
d
iv
i
d
e
alik
e
p
r
o
p
e
r
ties
th
r
o
u
g
h
n
eu
r
o
n
s
in
s
id
e
less
er
s
eq
u
en
tial
co
r
tex
th
at
ar
e
u
s
ed
in
f
av
o
r
o
f
o
b
ject
r
ec
o
g
n
it
io
n
in
s
id
e
m
an
d
r
ill
im
ag
e.
Featu
r
es
ar
e
r
eso
u
r
ce
f
u
lly
d
etec
ted
d
u
r
in
g
a
th
ea
tr
ic
al
f
ilter
ad
v
an
ce
s
o
as
to
id
en
tify
co
n
s
tan
t
p
o
in
ts
with
in
s
ize
s
p
ac
e.
I
m
a
g
e
k
e
y
s
ar
e
cr
ea
ted
to
war
d
p
er
m
it
o
n
b
eh
al
f
o
f
lim
ited
a
r
ith
m
etic
al
d
ef
o
r
m
atio
n
s
b
y
m
ea
n
s
o
f
r
ep
r
esen
tin
g
u
n
clea
r
p
ic
tu
r
e
g
r
ad
ien
ts
in
s
id
e
n
u
m
er
o
u
s
d
ir
ec
tio
n
p
lan
es
as
w
ell
as
o
n
n
u
m
e
r
o
u
s
s
ca
les.
T
h
e
k
ey
s
ar
e
u
tili
ze
d
b
ec
au
s
e
ef
f
o
r
t
to
war
d
a
n
ea
r
e
s
t
-
n
eig
h
b
o
r
in
d
e
x
in
g
tech
n
iq
u
e
to
war
d
id
e
n
tifie
s
ap
p
lican
t
o
b
jects
m
atch
es.
T
h
e
l
ast
co
n
f
ir
m
atio
n
o
f
ev
er
y
c
o
m
p
etitio
n
is
ac
h
i
ev
e
d
th
r
o
u
g
h
d
is
co
v
er
y
a
lo
w
-
r
esid
u
al
least
-
s
q
u
ar
es r
esu
lt in
ten
d
ed
o
n
b
e
h
alf
o
f
th
e
u
n
id
en
tifie
d
r
ep
r
o
d
u
ctio
n
p
ar
am
ete
r
s
.
T
h
e
d
etec
tio
n
p
e
r
f
o
r
m
an
ce
m
ig
h
t
b
e
ad
v
an
ce
en
h
an
c
ed
b
y
m
ea
n
s
o
f
ad
d
in
g
n
o
v
el
SIFT
ch
ar
ac
ter
is
tic
ty
p
es
to
war
d
s
lo
t
in
co
lo
r
,
tex
tu
r
e
,
alo
n
g
with
b
o
u
n
d
ar
y
g
r
o
u
p
in
g
,
also
u
n
r
el
iab
le
ch
ar
ac
ter
is
tic
s
izes
al
s
o
o
f
f
s
ets.
Scale
-
in
v
ar
ian
t
b
o
u
n
d
ar
y
g
r
o
u
p
in
g
to
w
ar
d
cr
ea
te
lim
ited
f
ig
u
r
e
-
g
r
o
u
n
d
d
is
cr
im
in
atio
n
s
[2
0
-
2
2
]
wo
u
l
d
ex
is
t
m
ain
ly
h
elp
f
u
l
o
n
o
b
ject
lim
itatio
n
s
w
h
er
ev
er
en
v
ir
o
n
m
en
t
d
is
o
r
d
e
r
b
e
ab
le
to
in
ter
f
er
e
with
ex
tr
a
f
ea
tu
r
es
.
T
h
e
in
d
e
x
in
g
with
co
n
f
ir
m
atio
n
s
tr
u
ctu
r
e
allo
ws
o
n
b
eh
alf
o
f
ea
c
h
an
d
ev
er
y
o
n
e
ty
p
e
o
f
lev
el
alo
n
g
with
alter
n
atio
n
in
v
ar
ian
t
f
ea
tu
r
es
to
war
d
s
u
r
v
iv
e
in
clu
d
e
d
in
ter
ested
in
a
p
ar
ticu
lar
m
o
d
el
illu
s
tr
atio
n
.
H
ig
h
est
s
tr
en
g
th
wo
u
ld
b
e
ac
h
ie
v
ed
th
r
o
u
g
h
d
etec
tin
g
a
lo
t
o
f
d
is
s
im
ilar
ch
ar
ac
ter
is
tic
ty
p
es
as
well
as
r
ely
in
g
u
n
d
er
t
o
p
o
f
th
e
in
d
e
x
in
g
as
well
as
clu
s
t
er
in
g
to
wa
r
d
c
h
o
o
s
e
in
d
iv
id
u
als
to
f
ac
ilit
ate
ar
e
m
ain
ly
h
elp
f
u
l w
ith
in
an
e
x
ac
t
in
g
p
ictu
r
e.
3.
P
RO
P
O
SE
D
WO
RK
T
h
e
p
lan
n
ed
FML
ND
tech
n
i
q
u
e
co
n
s
is
ts
o
f
th
r
ee
m
ajo
r
s
tag
es
:
p
r
e
-
p
r
o
ce
s
s
in
g
o
f
th
e
m
ag
n
etic
r
eso
n
an
ce
alo
n
g
with
co
m
p
u
t
ed
to
m
o
g
r
ap
h
y
im
ag
er
y
,
lear
n
in
g
o
f
th
e
lo
ca
l
d
escr
ip
t
o
r
s
,
as
well
as
ca
lcu
latio
n
o
f
th
e
p
r
e
d
icted
co
m
p
u
ted
to
m
o
g
r
ap
h
y
p
ictu
r
e
v
ia
f
ea
tu
r
e
m
atch
in
g
.
T
h
e
s
tag
es
o
f
t
h
e
p
l
an
n
ed
tech
n
iq
u
e
ar
e
ex
p
o
s
ed
with
in
Fig
u
r
e
1.
Fig
u
r
e
1
.
Pro
p
o
s
ed
p
r
o
ce
s
s
o
f
p
C
T
s
y
n
th
esis
f
r
o
m
MR d
ata
th
r
o
u
g
h
f
ea
t
u
r
e
m
atch
in
g
with
lear
n
ed
n
o
n
lin
ea
r
lo
ca
l d
escr
ip
to
r
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
N
o
va
l a
d
va
n
ce
n
o
n
-
lin
ea
r
d
es
crip
to
r
a
n
d
ch
a
r
a
cteris
tic
eq
u
iva
len
cy
to
…
(
S
o
w
mya
B
a
ch
u
)
49
3
.
1
.
P
re
-
pro
ce
s
s
inb
g
o
f
t
he
M
R
a
nd
C
T
i
m
a
g
er
y
T
h
e
p
ictu
r
e
d
ataset
u
tili
ze
d
d
u
r
in
g
th
is
ef
f
o
r
t
co
n
s
is
ts
o
f
T
1
w
alo
n
g
with
T
2
-
weig
ht
ed
(
T
2
w)
m
ag
n
etic
r
eso
n
an
ce
im
a
g
er
y
,
also
th
e
eq
u
iv
alen
t
c
o
m
p
u
te
d
to
m
o
g
r
ap
h
y
im
a
g
er
y
o
f
1
3
s
tr
o
n
g
p
atien
ts
.
T
h
e
ex
am
in
er
p
a
r
am
eter
o
f
th
e
im
a
g
er
y
ar
e
p
r
o
g
r
am
m
e
d
with
in
T
ab
le
1
.
T
o
war
d
b
e
g
in
th
r
o
u
g
h
,
th
e
N
-
4
b
ias
alter
atio
n
alg
o
r
ith
m
1
[
2
0
]
u
tili
ze
s
to
war
d
tak
in
g
awa
y
th
e
b
ias
p
astu
r
e
ar
tifa
ct
with
in
th
e
m
a
g
n
etic
r
eso
n
an
ce
im
ag
in
g
.
A
f
ter
th
at,
a
s
tr
en
g
th
n
o
r
m
aliza
tio
n
m
eth
o
d
[
2
1
]
is
u
s
ef
u
l
to
war
d
d
ec
r
ea
s
e
th
e
d
if
f
er
en
ce
d
iag
o
n
ally
th
e
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
o
f
d
is
s
im
ilar
p
atien
ts
.
T
h
e
in
ten
s
ities
o
f
th
e
m
ag
n
etic
r
eso
n
an
ce
im
ag
er
y
ar
e
e
x
ten
d
ed
to
war
d
v
ar
y
o
f
[
0
,
1
0
0
]
.
T
h
e
m
in
d
v
o
l
u
m
es
(
s
u
itab
le
im
ag
in
g
a
r
ea
)
ar
e
d
iv
id
ed
o
n
o
r
af
ter
th
e
C
T
s
ca
n
n
in
g
c
o
t
in
s
id
e
th
e
C
T
im
ag
er
y
th
r
o
u
g
h
th
r
esh
o
ld
in
g
.
L
astl
y
,
s
p
atial
n
o
r
m
aliza
tio
n
is
p
er
f
o
r
m
ed
v
ia
lin
ea
r
a
f
f
in
e
lis
tin
g
(
in
FLI
R
T
[
2
2
]
with
in
FS
L
2
b
y
cr
o
s
s
-
co
r
r
elatio
n
b
ec
au
s
e
th
e
p
ictu
r
e
c
o
m
p
ar
is
o
n
ass
ess
es)
to
war
d
lin
e
u
p
th
e
m
atch
in
g
MR
as
wel
l
as
co
m
p
u
ted
to
m
o
g
r
ap
h
y
v
o
lu
m
es
o
f
ev
er
y
to
ler
a
n
t.
T
h
ese
lin
ea
r
af
f
in
e
r
eg
is
ter
s
s
er
v
e
b
ec
a
u
s
e
of
th
e
s
o
u
r
ce
o
f
th
e
f
o
llo
win
g
s
tag
es
in
s
id
e
th
e
p
lan
n
ed
tech
n
iq
u
e.
T
ab
le
1
.
Scan
Par
am
eter
s
o
f
th
e
d
ataset
D
a
t
a
s
e
t
I
mag
i
n
g
p
a
r
a
met
e
r
s
M
R
I
S
c
a
n
n
e
r
:
G
E
M
e
d
i
c
a
l
S
y
s
t
e
m S
i
g
n
a
H
D
x
t
.
M
a
g
n
e
t
i
c
f
i
e
l
d
st
r
e
n
g
t
h
:
3
T.
S
l
i
c
e
t
h
i
c
k
n
e
ss:
2
.
5
m
m.
I
n
p
l
a
n
e
r
e
s
o
l
u
t
i
o
n
:
0
.
4
7
m
m
x
0
.
4
7
m
m.
T1
w
TR
=
2
8
8
4
.
7
4
ms,
TE=
7
.
9
ms
,
TI
=
9
6
0
m
s,
f
l
i
p
a
n
g
l
e
=
9
0
°
.
T1
w
TR
=
5
0
0
0
ms,
TE
=
1
0
0
.
4
ms,
f
l
i
p
a
n
g
l
e
=
9
0
°
.
S
c
a
n
n
e
r
:
G
E
M
e
d
i
c
a
l
S
y
s
t
e
ms
Li
g
h
t
S
p
e
e
d
P
r
o
1
6
.
S
l
i
c
e
t
h
i
c
k
n
e
ss:
2
.
5
m
m.
In
-
p
l
a
n
e
r
e
s
o
l
u
t
i
o
n
:
0
.
4
7
mm
x
0
.
4
7
mm
.
K
v
p
=
1
2
0
,
Ex
p
o
su
r
e
T
i
me=
8
0
0
ms,
X
-
r
a
y
T
u
b
e
c
u
r
r
e
n
t
=
3
0
0
mA
T
o
war
d
c
o
r
r
ec
tly
f
o
r
ec
ast
th
e
p
r
ed
icted
co
m
p
u
ted
t
o
m
o
g
r
a
p
h
y
im
a
g
er
y
th
r
o
u
g
h
m
ea
n
s
o
f
a
KNN
esti
m
ato
r
,
th
e
co
m
p
ar
is
o
n
o
th
er
wis
e
s
p
ac
e
am
o
n
g
th
e
f
ea
tu
r
es
m
u
s
t
r
ep
r
o
d
u
ce
o
r
else
s
u
r
v
iv
e
ass
o
ciate
d
with
th
e
co
m
p
ar
is
o
n
o
th
er
wis
e
s
p
ac
e
am
o
n
g
th
e
p
r
ed
icted
tar
g
ets.
U
s
u
ally
,
co
n
s
er
v
ativ
e
m
ag
n
etic
r
eso
n
an
c
e
im
ag
in
g
(
s
u
ch
as
T
1
w
as
well
as
T
2
w)
s
tr
en
g
th
ca
n
n
o
t
r
ep
r
o
d
u
ce
th
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ass
ess
m
en
t
o
p
en
ly
.
I
n
ad
d
itio
n
,
th
e
co
m
p
ar
is
o
n
am
o
n
g
th
e
r
aw
p
atch
e
s
o
f
a
m
ag
n
etic
r
eso
n
an
ce
p
i
ctu
r
e
m
ig
h
t
p
o
o
r
ly
r
ep
r
o
d
u
ce
th
e
co
m
p
ar
is
o
n
am
o
n
g
t
h
e
eq
u
iv
alen
t
c
o
m
p
u
ted
to
m
o
g
r
a
p
h
y
p
atch
es.
H
o
wev
e
r
,
th
e
h
an
k
ie
ty
p
e
’
s
co
n
tain
er
e
x
is
ts
r
ec
o
g
n
ize
d
t
h
r
o
u
g
h
t
h
e
s
tr
u
ctu
r
al
alo
n
g
with
r
elate
d
in
f
o
r
m
atio
n
o
f
h
u
g
e
s
p
atial
s
u
p
p
o
r
ts
in
s
id
e
th
e
m
a
g
n
etic
r
eso
n
an
c
e
im
ag
in
g
.
I
t
is
p
r
ef
er
r
ed
t
o
th
e
lim
ited
d
escr
ip
to
r
s
o
f
a
m
ag
n
e
tic
r
eso
n
an
ce
p
ictu
r
e
b
e
a
b
le
to
s
ig
n
if
y
t
h
e
s
tr
u
ctu
r
al
in
f
o
r
m
atio
n
also
ex
is
t u
s
e
to
war
d
class
if
y
ef
f
icien
t
ass
es
s
m
en
t
u
s
ed
o
n
b
eh
alf
o
f
KNN
wea
k
en
in
g
.
Su
p
er
v
is
ed
d
escr
ip
to
r
lear
n
in
g
(
SDL)
tech
n
iq
u
e
s
co
n
tain
er
is
u
tili
ze
d
to
r
ea
ch
th
is
tar
g
et.
3
.
2
.
L
ea
r
nin
g
o
f
t
he
lin
ea
r
des
cr
ipto
rs
SUR
F
's
d
etec
to
r
s
also
d
escr
ip
to
r
is
n
o
t
s
im
p
ly
q
u
ick
er
,
o
th
e
r
th
an
th
e
p
ast
b
e
to
o
f
u
r
th
e
r
r
ep
ea
tab
le
as
well
as
th
e
last
ex
tr
a
ch
ar
a
cter
is
tic
[
2
3
]
.
Hess
ian
-
b
ased
d
etec
to
r
s
ar
e
f
u
r
th
er
c
o
n
s
tan
t
a
s
well
as
r
ep
ea
tab
le
th
an
th
e
ir
Har
r
is
b
ased
co
u
n
ter
p
ar
t
alo
n
g
with
ex
p
er
im
e
n
tal
to
war
d
a
p
p
r
o
x
im
atio
n
s
i
m
ilar
to
th
e
Do
G
co
n
tain
er
ca
r
r
y
r
ate
o
n
a
s
m
all
r
ate
d
u
r
in
g
c
o
n
d
itio
n
s
o
f
lo
s
t
ex
ac
tn
ess
[
2
]
[
6
]
.
T
h
e
r
e
ar
e
m
ain
ly
two
s
tep
s
with
in
SUR
F:
3
.
2
.
1
.
I
nte
re
s
t
P
o
int
Det
ec
t
io
n
SUR
F
p
er
f
o
r
m
an
ce
n
ew
p
ict
u
r
e
to
war
d
in
te
g
r
al
p
ictu
r
e.
I
n
teg
r
al
p
ictu
r
e
w
h
ich
s
u
m
m
ed
r
eg
i
o
n
tab
les
is
a
m
id
way
illu
s
tr
atio
n
o
f
t
h
e
p
ictu
r
e
.
I
t
is
th
e
s
u
m
m
atio
n
o
f
in
ten
s
ity
ass
ess
m
en
t
s
o
f
th
e
en
tire
p
ix
els
with
in
co
n
tr
ib
u
tio
n
p
ictu
r
e.
I
m
ag
er
y
r
ec
tan
g
u
lar
ar
ea
s
h
ap
e
d
v
ia
s
o
u
r
ce
O=
(
0
,
0
)
alo
n
g
with
o
n
e
p
o
s
itio
n
X=
(
x
,
y
)
.
I
t g
iv
e
s
a
q
u
ick
ca
lc
u
latio
n
o
f
b
o
x
v
ar
iety
d
if
f
icu
lt
y
f
ilter
s
[
2
]
[
6
]
.
(
1
)
B
ased
o
n
to
p
o
f
in
teg
r
al
im
a
g
e,
th
er
e
ar
e
s
im
p
ly
th
r
ee
o
p
er
atio
n
s
(
ad
d
itio
n
o
th
er
wis
e
s
u
b
tr
ac
tio
n
)
n
ec
ess
itate
o
n
b
eh
al
f
o
f
s
ch
e
m
in
g
ca
lcu
latio
n
o
f
in
te
n
s
ity
ass
es
s
m
en
t
s
o
f
p
ix
els
m
o
r
e
i
n
th
e
least
v
er
tical
r
ec
tan
g
u
lar
r
eg
io
n
.
I
n
te
g
r
al
im
ag
e
as sh
o
wn
in
Fi
g
u
r
e
2
.
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.
9
,
No
.
1
,
Ap
r
il
20
20
:
4
6
–
5
6
50
Fig
u
r
e
2
.
Usi
n
g
in
te
g
r
al
im
ag
e
r
y
,
it tak
es o
n
l
y
th
r
ee
a
d
d
itio
n
s
also
f
o
u
r
m
em
o
r
y
ac
ce
s
s
es to
war
d
ca
lcu
late
th
e
s
u
m
o
f
in
ten
s
ities
in
s
id
e
a
r
ec
t
an
g
u
lar
r
eg
io
n
o
f
a
n
y
s
ize
I
n
teg
r
al
p
ictu
r
e
is
co
m
p
licated
th
r
o
u
g
h
b
o
x
f
ilter
.
B
o
x
f
ilter
is
esti
m
at
ed
f
ilter
o
f
Gau
s
s
ian
f
ilter
.
T
h
e
Hess
ian
m
ed
iu
m
ℋ
(
X,
σ
)
,
(
s
in
ce
(
2
)
)
wh
er
ev
er
X=
(
x
,
y
)
o
f
an
p
ictu
r
e
I
,
o
n
lev
el
σ
is
d
ef
in
ite
as
:
(
2
)
W
h
er
e
(
,
)
(
L
a
p
lacia
n
o
f
Ga
u
s
s
ian
)
b
e
co
n
v
o
lu
tio
n
o
f
t
h
e
Gau
s
s
ian
n
ex
t
ar
r
an
g
e
d
er
i
v
e
d
2
2
(
)
th
r
o
u
gh
(
2
)
with
in
p
o
s
itio
n
X
also
lik
ewise
m
ea
n
t o
n
b
eh
alf
o
f
(
,
)
alo
n
g
with
(
,
)
.
3
.
2
.
2
.
I
nte
re
s
t
P
o
int
Descript
io
n
W
ith
in
o
r
d
er
to
war
d
in
v
a
r
ian
t
to
war
d
im
ag
e
r
o
tatio
n
,
we
r
ec
o
g
n
ize
a
r
ep
r
o
d
u
ci
b
le
d
ir
ec
ti
o
n
u
s
ed
o
n
b
eh
alf
o
f
th
e
n
o
tice
p
o
s
itio
n
s
B
ec
au
s
e
o
f
th
at,
i
n
itial
co
m
p
u
te
Haa
r
wav
elet
r
ea
ctio
n
s
with
in
x
alo
n
g
with
y
p
ath
in
s
id
e
a
r
o
u
n
d
ar
ea
o
f
r
ad
iu
s
6
s
a
p
p
r
o
x
im
ately
t
h
e
n
o
tice
p
o
s
itio
n
,
th
r
o
u
g
h
s
ca
l
e
s
(
s
am
p
lin
g
s
tep
d
ep
en
d
s
o
n
s
)
u
n
d
e
r
wh
ich
th
e
n
o
tice
p
o
s
itio
n
b
e
d
etec
t
c
an
b
e
s
ee
n
at
Fig
u
r
e
3
.
T
h
e
d
i
m
en
s
io
n
o
f
wav
ele
t
wh
ich
is
weig
h
in
g
m
ac
h
in
e
d
ep
en
d
ed
alo
n
g
with
its
s
id
e
len
g
th
is
4
s
.
T
o
wa
r
d
ca
lcu
late
t
h
e
an
s
wer
with
in
x
o
r
else y
r
o
u
te
o
n
o
n
e
s
ize
s
im
p
ly
s
ix
o
p
er
atio
n
s
ar
e
n
ee
d
ed
[
24
].
P
r
ev
io
u
s
ly
th
e
wav
elet
r
ea
ctio
n
s
a
r
e
d
esig
n
ed
as
well
as
wei
g
h
ted
th
r
o
u
g
h
a
Gau
s
s
ian
σ
=2
s
ce
n
ter
ed
o
n
th
e
n
o
tice
p
o
s
itio
n
.
T
h
e
r
ea
ctio
n
s
ar
e
r
ep
r
esen
tin
g
s
in
c
e
tip
s
with
in
a
g
ap
th
r
o
u
g
h
t
h
e
s
tr
aig
h
t
r
ea
ctio
n
f
o
r
ce
n
ex
t
t
o
war
d
t
h
e
a
b
s
ciss
a
as
well
as
th
e
p
er
p
e
n
d
icu
lar
r
ea
ctio
n
f
o
r
ce
b
esid
e
th
e
o
r
d
i
n
ate.
Fin
d
m
a
x
im
u
m
th
e
am
o
u
n
t
o
f
ev
e
r
y
r
esp
o
n
s
e
wh
ich
is
wav
elet
r
ea
ctio
n
wi
th
in
ev
er
y
d
o
wn
tr
a
n
s
o
m
(
π/3
tr
an
s
o
m
d
ir
ec
tio
n
)
(
o
b
s
er
v
e
Fig
u
r
e
4
)
.
T
h
e
f
lat
,
as
well
as
u
p
r
ig
h
t
r
esp
o
n
s
es
i
n
s
id
e
th
e
win
d
o
w,
ar
e
s
u
m
m
ed
.
Fro
m
th
ese
two
h
o
r
izo
n
tals
as
well
as
v
er
tical,
s
u
m
m
ed
r
esp
o
n
s
es
n
ex
t
g
i
v
e
way
a
lim
ited
d
ir
ec
tio
n
v
ec
to
r
.
T
h
e
d
ir
ec
tio
n
o
f
th
e
n
o
tice
p
o
s
itio
n
co
n
tain
er
e
x
is
ts
d
ef
in
ite
th
r
o
u
g
h
r
esu
lt b
e
s
t su
ch
v
ec
to
r
m
o
r
e
th
an
ev
er
y
tr
an
s
o
m
.
Fig
u
r
e
3
.
Haa
r
wa
v
elet
f
ilter
s
to
war
d
ca
lcu
late
h
r
r
esp
o
n
s
e
with
in
x
(
lef
t)
as we
ll a
s
y
d
i
r
ec
tio
n
(
r
ig
h
t)
.
T
h
e
s
h
ad
y
p
ar
ts
co
n
tain
t
h
e
w
eig
h
t
-
1
alo
n
g
with
th
e
lig
h
t p
ar
ts
+1
Fig
u
r
e
4
.
Or
ien
tatio
n
ass
ig
n
m
en
t: A
Sli
d
in
g
o
r
ien
tatio
n
win
d
o
e
o
f
s
ize
/
3
d
etec
ts
th
e
d
o
m
in
a
n
t o
r
ien
tatio
n
o
f
th
e
Gau
s
s
in
weig
h
ted
Haa
r
wav
elet
r
esp
o
n
s
es a
t e
v
er
y
s
a
m
p
le
p
o
in
t w
ith
in
a
cir
cu
lar
n
eig
h
b
o
r
h
o
o
d
ar
o
u
n
d
th
e
in
ter
est p
o
in
t.
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
N
o
va
l a
d
va
n
ce
n
o
n
-
lin
ea
r
d
es
crip
to
r
a
n
d
ch
a
r
a
cteris
tic
eq
u
iva
len
cy
to
…
(
S
o
w
mya
B
a
ch
u
)
51
T
o
war
d
ex
tr
ac
t
,
th
e
d
escr
ip
t
o
r
,
cu
b
e
ar
ea
wh
ic
h
d
im
en
s
i
o
n
is
2
0
s
ar
e
c
o
n
s
tr
u
cted
u
n
d
er
to
p
o
f
in
ter
ested
p
o
in
ts
.
E
x
am
p
le
s
o
f
s
u
ch
s
q
u
ar
e
r
e
g
io
n
s
ar
e
ill
u
s
tr
ate
d
in
Fig
u
r
e
5.
Fig
u
r
e
5
.
Deta
il o
f
th
e
Gr
af
f
iti
s
ce
n
e
s
h
o
win
g
th
e
s
ize
o
f
th
e
o
r
ien
ted
d
escr
ip
t
o
r
win
d
o
w
at
d
if
f
er
en
t
T
h
e
wav
elet
r
esp
o
n
s
es
d
x
with
d
y
ar
e
s
u
m
m
ed
u
p
m
o
r
e
th
an
ev
er
y
s
u
b
-
r
eg
io
n
alo
n
g
with
f
o
r
m
a
f
ir
s
t
s
et
o
f
e
n
tr
ies
in
s
id
e
th
e
f
ea
tu
r
e
v
ec
to
r
.
D
u
r
in
g
o
r
d
er
t
o
war
d
b
r
in
g
in
in
f
o
r
m
atio
n
c
o
n
ce
r
n
in
g
th
e
p
o
lar
ity
o
f
th
e
in
ten
s
ity
ch
an
g
es,
ex
tr
ac
t
th
e
s
u
m
o
f
th
e
ab
s
o
lu
te
ass
ess
m
en
t
s
o
f
th
e
r
esp
o
n
s
es,
|
d
x
|
alo
n
g
with
|
d
y
|
,
ea
ch
s
u
b
-
r
eg
io
n
h
av
e
a
f
o
u
r
-
d
im
en
s
io
n
al
d
escr
ip
to
r
v
ec
to
r
V
u
s
e
d
f
o
r
its
u
n
d
e
r
ly
in
g
in
ten
s
ity
s
tr
u
ctu
r
e
V=
(
∑d
x
,
∑d
y
,
∑
|
d
x
|
,
∑
|
d
x
|
)
.
C
o
n
ca
ten
atin
g
th
is
u
s
ed
f
o
r
all,
4
x
4
s
u
b
-
r
eg
io
n
s
,
with
th
ese
r
esu
lt
s
with
in
a
d
escr
ip
to
r
v
ec
to
r
o
f
len
g
th
is
6
4
.
T
h
e
w
av
elet
r
esp
o
n
s
es
ar
e
in
v
a
r
ian
t
to
war
d
a
b
ias
in
s
id
e
illu
m
in
a
tio
n
(
o
f
f
s
et)
as
well
as I
n
v
ar
ian
ce
to
wa
r
d
co
n
tr
ast (
a
s
ca
le
f
ac
to
r
)
is
ac
h
iev
e
d
th
r
o
u
g
h
t
u
r
n
in
g
th
e
d
escr
ip
to
r
in
t
o
a
u
n
it
v
ec
to
r
.
3.
3
.
L
ea
rning
o
f
t
he
no
nli
nea
r
des
cr
ipto
rs
T
h
e
lear
n
ed
d
escr
ip
to
r
(
lin
ea
r
d
escr
ip
to
r
)
with
in
(
3
)
is
p
r
esen
tly
th
e
lin
ea
r
em
b
ed
d
i
n
g
o
f
th
e
m
o
s
t
im
p
o
r
tan
t
d
escr
ip
to
r
.
T
o
war
d
g
et
f
u
r
th
er
d
o
m
in
an
t
d
escr
ip
t
o
r
;
f
u
r
th
er
e
x
p
an
d
th
e
lin
ea
r
f
o
r
m
u
latio
n
with
in
.
(
1
)
As
well
as
(
3
)
in
ter
ested
in
th
e
n
on
-
lin
ea
r
f
o
r
m
u
latio
n
s
b
y
th
e
k
er
n
el
m
eth
o
d
.
T
h
e
m
ain
d
escr
ip
to
r
is
p
r
ed
ictab
le
in
ter
ested
in
a
h
i
g
h
-
d
im
e
n
s
io
n
al
g
ap
t
h
r
o
u
g
h
a
n
o
n
-
lin
ea
r
m
ap
Φ
(
)
.
T
h
e
ch
a
r
a
cter
is
tic
p
lan
Φ
(
)
o
f
th
e
p
o
s
itiv
e
d
ef
in
ite
(
PD)
ess
en
tial
p
ar
t
(
,
′
)
[
2
5
]
m
a
p
s
th
e
l
in
ea
r
d
e
s
cr
ip
to
r
d
in
ter
ested
i
n
th
e
Hilb
er
t g
ap
[
2
6
]
th
r
o
u
g
h
lin
ea
r
in
ter
n
al
cr
ea
tio
n
<.
>,
i
.
e.
(
3
)
T
o
war
d
r
ed
u
ce
in
th
e
d
ir
ec
t
io
n
o
f
b
ar
e
-
b
o
n
es
o
p
tim
izatio
n
d
if
f
icu
lty
;
we
tak
e
u
n
d
er
th
e
o
p
en
ch
ar
ac
ter
is
tic
p
lan
tech
n
iq
u
e
to
war
d
esti
m
ated
a
k
n
o
w
n
p
o
s
itiv
e
d
ef
i
n
ite
k
er
n
el
(
,
′
)
th
r
o
u
g
h
a
ch
ar
ac
ter
is
tic
m
ap
Φ
(
)
:
d
⟶
.
T
h
e
attain
ed
p
r
o
tu
b
er
an
ce
co
n
tain
er
is
ap
p
r
o
x
im
ate
d
th
r
o
u
g
h
lin
ea
r
en
h
an
ce
d
s
u
p
er
v
is
ed
d
escr
ip
t
o
r
lear
n
in
g
tech
n
iq
u
e.
P
r
ec
is
e
ch
ar
ac
ter
is
tic
m
ap
s
b
e
a
p
p
r
o
p
r
iate
b
el
o
w
th
e
r
estra
in
t
to
ju
s
t
if
a
k
n
o
wn
n
o
n
-
E
u
clid
ea
n
m
etr
ic
b
e
p
r
eser
v
ativ
e
alo
n
g
with
s
tan
d
ar
d
ize
d
.
T
h
e
p
r
eser
v
ativ
e
k
er
n
el
(
,
′
)
o
n
to
p
o
f
R
q
co
n
tain
er
is
ex
p
r
ess
ed
as:
(
4
)
T
h
e
s
tan
d
ar
d
ized
k
er
n
el
co
n
t
ain
er
m
o
r
eo
v
er
b
e
ex
p
r
ess
ed
s
in
ce
(
,
′
)
=
(
,
′
)
m
ea
n
t
o
n
b
eh
alf
o
f
ev
er
y
>0
.
F
r
eq
u
en
tly
u
s
ed
k
er
n
el
[
2
7
-
2
9
]
co
m
p
r
is
es th
e
s
u
b
s
eq
u
en
t:
2
k
er
n
el:
(
5
)
J
en
s
en
-
Sh
an
n
o
n
(
J
S)
k
er
n
el:
(
6
)
I
n
ter
s
ec
tio
n
k
er
n
el:
(
7
)
T
h
e
o
p
e
n
f
ea
t
u
r
e
m
a
p
m
eth
o
d
ap
p
r
o
x
im
ates
p
r
o
tu
b
er
a
n
ce
Φ
(
)
b
y
s
ep
ar
ate
Fo
u
r
ier
ch
a
n
g
e
t
h
r
o
u
g
h
a
s
am
p
le
r
ate
r
wh
ile
Φ
̂
(
)
.
Ma
in
ly
th
ick
SUR
F
d
escr
ip
to
r
s
ar
e
p
r
ed
ictab
le
n
o
n
-
lin
ea
r
ly
in
te
r
e
s
ted
in
a
h
ig
h
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.
9
,
No
.
1
,
Ap
r
il
20
20
:
4
6
–
5
6
52
d
im
en
s
io
n
al
g
a
p
(
2
+
1
)
×
1
b
ec
au
s
e
th
e
o
p
en
ch
ar
ac
ter
is
tic
m
ap
Φ
̂
(
)
.
E
v
e
r
y
p
r
e
d
ictab
le
m
ain
d
en
s
e
SU
R
F
d
escr
ip
to
r
is
n
ex
t
r
ea
r
r
an
g
e
d
in
ter
ested
in
a
n
M
×N
m
atr
i
x
as
well
as
jo
in
t
th
r
o
u
g
h
th
e
r
aw
p
atch
s
in
ce
a
n
o
n
lin
ea
r
d
escr
ip
t
o
r
.
An
alo
g
o
u
s
ly
to
war
d
co
n
tr
o
l
lear
n
o
f
li
n
ea
r
d
escr
ip
to
r
,
we
u
tili
ze
th
e
s
im
ilar
f
o
r
m
u
latio
n
with
in
(
1
)
to
war
d
d
is
co
v
er
th
e
in
v
ar
ian
t
m
ed
iu
m
W
alo
n
g
with
th
e
d
i
s
cr
im
in
ativ
e
m
ed
iu
m
V
u
s
ed
o
n
b
eh
alf
o
f
th
e
n
o
n
lin
ea
r
h
ig
h
-
d
im
en
s
i
o
n
al
d
escr
ip
to
r
Φ
̂
(
)
.
A
f
in
al
L
ND
is
o
b
tain
ed
as
′
=
Φ
̂
(
)
.
Fig
u
r
e
6
.
(
a
)
,
as
well
as
(
b
)
h
er
e,
s
ig
n
if
y
co
m
p
lete
f
au
lt
o
f
m
ed
iu
m
W
alo
n
g
with
V
co
n
s
eq
u
en
ce
s
u
s
ed
o
n
b
eh
alf
o
f
E
q
u
ati
on
(
1
)
.
T
h
e
d
is
s
im
ilar
ity
with
in
p
er
p
en
d
icu
lar
alig
n
m
en
t
is
in
d
ica
tin
g
co
m
p
lete
f
au
lt
am
o
n
g
th
e
p
r
esen
t
as
well
as
b
ef
o
r
e
th
e
s
o
l
u
tio
n
t
h
r
o
u
g
h
th
e
iter
atio
n
s
.
Ma
tr
ices
W
,
as
well
as
V,
r
em
ain
s
u
n
af
f
ec
ted
f
o
llo
win
g
th
e
th
i
r
d
iter
atio
n
,
th
er
ef
o
r
e
r
ep
r
es
en
tin
g
th
e
b
est
ex
p
lan
atio
n
o
f
E
q
.
(
1
)
h
a
v
e
co
n
v
er
g
ed
.
T
o
war
d
u
n
d
o
u
b
te
d
ly
clar
if
y
th
e
th
o
u
g
h
t
o
f
lear
n
ed
n
o
n
-
lin
ea
r
d
escr
ip
to
r
,
we
s
u
p
p
ly
th
e
p
s
eu
d
o
-
co
d
e
u
s
ed
in
f
av
o
r
o
f
th
e
lear
n
in
g
p
r
o
ce
s
s
with
in
p
r
o
ce
d
u
r
e
1
.
(
a)
(
b
)
Fig
u
r
e
6
.
Me
an
ab
s
o
lu
te
er
r
o
r
o
f
d
is
cr
im
in
ativ
e
m
atr
i
x
W
alo
n
g
with
V
am
o
n
g
t
h
e
p
r
esen
t
o
u
tco
m
e
also
th
e
p
r
ev
io
u
s
r
esu
lt.
a
)
R
esu
lt o
f
lin
ea
r
d
escr
ip
to
r
s
.
b
)
R
esu
lt o
f
n
o
n
lin
ea
r
d
escr
ip
to
r
s
.
Alg
o
rit
hm
1
:
L
ea
rning
o
f
t
h
e
no
nli
nea
r
lo
ca
l descript
o
rs
I
np
ut:
P
r
ep
ar
atio
n
p
air
o
f
th
e
m
ag
n
etic
r
eso
n
a
n
ce
alo
n
g
with
co
m
p
u
ted
r
eso
n
an
ce
im
ag
e
r
y
.
O
utput
:
Dis
cr
im
in
ativ
e
m
ed
iu
m
W
alo
n
g
with
V.
Sta
g
e
-
1:
A
s
s
u
m
e
in
ten
s
e
SUR
F
to
war
d
tak
in
g
o
u
t
th
e
e
x
ten
s
iv
e
v
ar
iety
o
f
f
ea
tu
r
e
s
as
well
as
th
e
s
tr
u
ctu
r
al
in
f
o
r
m
atio
n
o
f
t
h
e
M
R
im
ag
er
y
.
Sta
g
e
-
2:
P
lan
th
e
e
x
tr
ac
t
in
te
n
s
e
SUR
F
d
escr
ip
to
r
s
in
ter
est
ed
in
th
e
h
ig
h
d
im
en
s
io
n
al
s
p
ac
e
b
y
o
p
e
n
f
ea
tu
r
e
m
ap
s
,
alo
n
g
with
s
u
b
s
eq
u
en
t
ly
th
ey
o
b
tain
d
escr
ip
to
r
s
in
clu
d
ed
th
r
o
u
g
h
r
aw
p
atch
s
tr
en
g
th
in
d
icate
th
e
n
o
n
lin
ea
r
d
escr
ip
to
r
s
.
Sta
g
e
-
3:
G
ath
er
a
b
u
lk
y
am
o
u
n
t
o
f
MRI
n
o
n
lin
ea
r
d
escr
ip
to
r
p
atch
es
alo
n
g
with
th
eir
eq
u
i
v
alen
t
C
T
s
tr
en
g
th
p
atch
es.
Sta
g
e
-
4:
E
x
p
l
o
r
e
C
ad
jace
n
t n
eig
h
b
o
r
s
o
f
ev
e
r
y
C
T
ar
ea
i
n
s
id
e
a
co
n
s
tr
ictio
n
s
p
atial
v
ar
iet
y
to
war
d
attain
th
e
m
ed
iu
m
S.
Sta
g
e
-
5:
I
ter
ativ
ely
ex
p
lain
th
e
b
est
r
o
le
E
q
u
atio
n
(
1
)
t
o
d
is
co
v
er
th
e
b
est
ef
f
ec
t
also
f
in
d
th
e
d
is
cr
im
in
ativ
e
m
atr
ices W alo
n
g
with
V.
3
.
2
.
2
.
Ca
lcula
t
io
n o
f
t
he
predict
ed
co
m
pu
t
ed
t
o
m
o
g
ra
ph
y
pict
ure
t
hro
ug
h f
ea
t
ure
ma
t
ching
D
esig
n
ed
o
n
b
e
h
alf
o
f
e
v
er
y
p
o
s
itio
n
x
⊂
C
i
n
s
id
e
in
p
u
t
T
1
w
as
well
a
s
T
2
w
m
ag
n
etic
r
eso
n
an
ce
I
m
ag
er
y
,
r
e
d
u
ce
th
e
r
ate
p
u
r
p
o
s
e
to
war
d
ap
p
r
o
x
im
atio
n
t
h
e
eq
u
iv
alen
t
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ar
ea
f
(
x
)
ce
n
ter
ed
o
n
x
:
(
8
)
W
h
er
e
̂
(
)
an
esti
m
ato
r
alo
n
g
with
C
b
e
th
e
v
o
x
el
lo
ca
te
in
s
id
e
s
u
itab
le
p
ictu
r
e
ar
ea
.
K
-
N
ea
r
est
Neig
h
b
o
r
wea
k
en
in
g
is
u
tili
zin
g
to
war
d
g
u
ess
th
e
ass
ess
m
e
n
t
o
f
p
u
r
p
o
s
e
̂
(
)
.
k
n
ea
r
est
n
eig
h
b
o
r
s
b
e
s
ea
r
ch
as
well
as
ch
o
s
en
in
a
p
e
r
m
a
n
en
t
s
p
atial
v
ar
iety
o
f
e
v
er
y
p
o
s
itio
n
x
with
in
th
e
m
ag
n
etic
r
eso
n
an
ce
im
ag
e
r
y
u
s
ed
o
n
b
eh
alf
o
f
th
e
lear
n
lin
ea
r
o
th
e
r
wis
e
n
o
n
-
lin
ea
r
d
escr
ip
to
r
d
x
.
T
h
e
k
n
ea
r
est
n
eig
h
b
o
r
s
o
f
d
x
with
i
n
M
ag
n
etic
r
eso
n
an
ce
im
ag
e
r
y
alo
n
g
with
eq
u
iv
ale
n
t
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ar
ea
s
co
n
tain
er
is
d
en
o
te
d
as
{
,
(
=
1
,
2
,
…
)
}
.
D
u
r
in
g
th
e
d
iv
er
s
e
o
f
th
e
m
ag
n
etic
r
eso
n
an
ce
d
escr
ip
t
o
r
s
,
a
lin
ea
r
esti
m
ate
d
x
co
n
tain
er
is
o
b
tain
e
d
b
y
:
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
N
o
va
l a
d
va
n
ce
n
o
n
-
lin
ea
r
d
es
crip
to
r
a
n
d
ch
a
r
a
cteris
tic
eq
u
iva
len
cy
to
…
(
S
o
w
mya
B
a
ch
u
)
53
(
9
)
T
h
e
b
iased
co
ef
f
icien
t
v
ec
to
r
w
b
e
u
tili
ze
d
to
war
d
ca
lcu
la
t
in
g
th
e
q
u
an
tity
o
f
ass
ess
m
e
n
t
am
o
n
g
d
escr
ip
to
r
d
x
with
it
s
k
ad
jace
n
t
n
eig
h
b
o
r
s
.
T
h
e
b
iased
c
o
ef
f
icien
t
v
ec
to
r
w
c
o
n
tain
er
is
co
n
s
id
er
ed
b
asically
b
y
a
Gau
s
s
ian
k
er
n
el
f
u
n
cti
o
n
:
(
1
0
)
is
d
eter
m
in
ed
b
y
th
e
f
o
llo
win
g
:
(
1
1
)
W
h
er
e
(
)
is
th
e
q
th
n
ea
r
est
n
eig
h
b
o
r
o
f
d
x
.
D
u
r
in
g
t
h
e
ex
p
er
i
m
en
ts
,
p
ar
am
eter
u
s
u
ally
wo
r
k
in
g
s
f
in
e
wh
en
q
=4
.
O
n
ce
o
b
tain
t
h
e
weig
h
ted
c
o
ef
f
icien
ts
w
,
th
e
C
T
v
ec
to
r
o
r
else
ar
ea
u
s
ed
o
n
b
eh
al
f
o
f
e
v
er
y
MR d
escr
ip
to
r
d
x
co
n
tain
er
ar
e
esti
m
ated
as:
(
1
2
)
W
ith
th
e
eq
u
iv
ale
n
t
,
C
T
p
atc
h
es
DC
T
k
alo
n
g
with
t
h
e
w
eig
h
ted
c
o
ef
f
icien
t
v
ec
to
r
w
.
F
o
llo
win
g
ev
er
y
o
n
e
o
f
th
e
C
T
p
atc
h
es
u
s
ed
o
n
b
e
h
alf
o
f
an
i
n
p
u
t
MR
p
ictu
r
e
is
p
r
ed
ictab
le,
a
b
iased
ty
p
ical
p
r
o
ce
s
s
is
p
er
f
o
r
m
ed
u
n
d
er
to
p
o
f
th
e
o
v
er
lap
p
ed
C
T
p
atch
es to
war
d
g
et
t
h
e
last
p
r
ed
ict
p
C
T
p
ictu
r
e.
3.
4
.
Sim
ula
t
i
o
n o
f
AC
U
s
u
ally
,
to
g
eth
er
with
th
e
m
ag
n
etic
r
eso
n
an
ce
,
as
well
as
C
T
im
ag
er
y
u
s
ed
o
n
b
e
h
alf
o
f
o
n
e
p
atien
t,
is
s
im
p
ly
ac
q
u
ir
ed
with
in
th
e
h
ea
lth
ce
n
ter
.
T
h
o
u
g
h
,
MRI,
C
T
also
PET
im
ag
er
y
m
ea
n
t
in
f
av
o
r
o
f
s
in
g
l
e
p
atien
t
is
h
ar
d
ly
ev
e
r
ev
er
y
o
b
tain
ab
le.
Ho
f
m
a
n
n
et
al.
[
4
]
p
lan
n
ed
t
o
m
ak
e
u
s
e
o
f
a
r
e
p
licated
u
s
u
al
PET
p
ictu
r
e
to
esti
m
ate
th
e
co
llis
io
n
o
f
p
C
T
atten
u
atio
n
co
r
r
ec
tio
n
d
u
r
i
n
g
th
e
ev
alu
atio
n
o
f
ac
cu
r
ate
co
m
p
u
te
d
to
m
o
g
r
a
p
h
y
atten
u
atio
n
c
o
r
r
ec
tio
n
.
S
u
b
s
eq
u
en
t
th
e
a
d
v
an
ce
u
s
ed
,
u
tili
ze
th
e
Po
s
itro
n
em
is
s
io
n
to
m
o
g
r
a
p
h
y
/M
ag
n
etic
R
eso
n
an
ce
I
m
ag
in
g
(
MN
I
1
5
2
T
1
w)
tem
p
late3
to
war
d
r
ep
r
o
d
u
ce
t
h
e
lo
s
t
PET
r
ec
o
r
d
s
to
war
d
ass
ess
in
g
th
e
p
r
esen
tatio
n
o
f
atten
u
atio
n
co
r
r
ec
tio
n
[
2
2
]
.
W
e
in
itial
lin
e
u
p
th
e
m
ag
n
etic
r
eso
n
an
ce
im
ag
in
g
p
atter
n
to
war
d
th
e
m
ag
n
etic
r
eso
n
an
ce
im
ag
i
n
g
o
f
ev
er
y
is
s
u
e
p
atien
t
b
y
th
e
d
ef
o
r
m
a
b
le
r
eg
is
ter
to
o
lb
o
x
FNI
R
T
4
with
in
th
e
F
SL
2
.
T
h
e
o
b
tain
ed
tr
an
s
f
o
r
m
a
tio
n
s
ar
e
n
ex
t
u
s
ef
u
l
to
wa
r
d
th
e
18
F
-
FDG
PET
tem
p
late
p
ictu
r
e
to
war
d
g
et
th
e
s
im
u
lated
PET
im
ag
er
y
.
W
ith
in
[
7
]
,
t
h
e
atten
u
atio
n
a
s
s
es
s
m
en
t
s
wer
e
tr
an
s
f
o
r
m
ed
lik
e
o
f
Ho
u
n
s
f
ield
u
n
its
(
HU)
to
war
d
L
AC
s
b
y
5
1
1
k
eV
with
in
cm
-
1
.
R
ed
u
ctio
n
m
ap
(
µ
-
m
ap
)
is
co
m
p
u
ted
v
ia
u
s
in
g
a
p
iece
-
w
is
e
lin
ea
r
m
ap
[
3
0
]
co
n
v
er
s
io
n
; t
h
e
co
m
p
u
ted
to
m
o
g
r
ap
h
y
ass
ess
m
en
t
s
b
e
tr
an
s
f
o
r
m
ed
in
te
r
ested
in
L
AC
s
lik
e
f
o
llo
w:
(
1
3
)
W
h
er
e
I
CT
d
e
n
o
te
th
e
C
T
p
i
ctu
r
e
in
d
icat
o
r
s
tr
en
g
th
,
e=
9
.
6
×1
0
-
5
cm
-
1
,
as
well
as
a
i
s
a
s
tab
le
d
ep
en
d
i
n
g
u
n
d
er
X
-
r
ay
tu
b
e
v
o
ltag
es.
T
h
e
µ
-
m
ap
r
esu
ltin
g
s
tar
t
th
e
ac
cu
r
ate
co
m
p
u
ted
to
m
o
g
r
ap
h
y
as
well
as
p
C
T
is
p
r
im
ar
y
p
r
ed
ictab
le
to
war
d
o
b
tain
s
th
e
r
e
d
u
ctio
n
r
ec
o
r
d
s
of
PET
pCT
b
y
t
h
e
R
ad
o
n
alo
n
g
with
co
n
v
er
s
e
R
ad
o
n
tr
an
s
f
o
r
m
atio
n
s
.
T
o
war
d
g
et
PET
pCT
,
th
e
u
n
c
o
r
r
ec
ted
s
o
n
o
g
r
am
r
ec
o
r
d
s
u
s
ed
f
o
r
atten
u
atio
n
ar
e
r
eq
u
ir
e
d
:
(
1
4
)
W
h
er
e
R
a
d
en
o
te
R
ad
o
n
tr
a
n
s
f
o
r
m
.
Nex
t,
PET
pCT
is
r
ec
o
n
s
tr
u
cted
b
y
th
e
co
r
r
ec
ted
s
o
n
o
g
r
a
m
o
n
ce
Sin
is
o
b
tain
ed
:
(
1
5
)
w
h
er
ev
er
R
a
-
1
b
e
th
e
c
o
n
v
e
r
s
e
R
ad
o
n
tr
an
s
f
o
r
m
s
.
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.
9
,
No
.
1
,
Ap
r
il
20
20
:
4
6
–
5
6
54
3.
4
.
E
v
a
lua
t
i
o
n m
ea
s
ures
W
e
esti
m
ate
th
e
m
eth
o
d
u
s
ed
in
f
a
v
o
r
o
f
p
r
e
d
ict
p
C
T
im
ag
er
y
with
ca
s
e
-
wis
e
leav
e
-
one
-
o
u
t
cr
o
s
s
-
v
alid
atio
n
(
L
OOCV).
Su
b
s
eq
u
en
t
L
ad
e
f
o
g
e
d
et
al.
[
3
1
]
,
f
o
u
r
q
u
an
titativ
e
p
r
o
ce
d
u
r
es
b
e
wo
r
k
in
g
to
war
d
esti
m
ate
th
e
p
r
esen
tatio
n
o
f
t
h
e
tech
n
iq
u
es.
−
M
ea
n a
bs
o
lute
er
ro
r
(
M
AE
)
:
Me
an
ab
s
o
lu
te
e
r
r
o
r
p
r
o
ce
d
u
r
es th
e
v
o
x
el
-
wis
e
e
r
r
o
r
(
with
i
n
HU)
,
wh
ich
co
n
tain
er
ex
is
t f
o
r
m
u
late
s
in
c
e
f
o
llo
w:
(
1
6
)
W
h
er
ev
er
C
b
e
th
e
v
o
x
el
lo
ca
t
e
with
in
a
s
u
itab
le
p
ictu
r
e
ar
ea
,
also
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
a
lo
n
g
with
p
r
ed
icted
co
m
p
u
ted
to
m
o
g
r
ap
h
y
in
d
icate
th
e
ac
cu
r
ate
C
T
alo
n
g
with
th
e
p
r
e
d
ict
p
C
T
p
ictu
r
e,
co
r
r
esp
o
n
d
in
g
ly
.
−
P
ea
k
s
ig
na
l
-
to
-
no
is
e
ra
t
io
(
P
SNR)
:
Peak
s
ig
n
al
t
o
n
o
is
e
r
at
io
(
with
in
d
B
)
is
d
ef
in
e
d
lik
e
f
o
llo
w:
(
1
7
)
W
h
er
e
Q
d
en
o
te
th
e
h
ig
h
est
ass
es
s
m
en
t
o
f
in
ten
s
ity
with
in
th
e
p
r
o
p
er
ly
co
m
p
u
ted
t
o
m
o
g
r
ap
h
y
as
well
as p
C
T
.
4.
E
XP
E
R
I
M
E
N
T
A
L
RE
SUL
T
S
T
h
e
r
esu
lt o
f
t
h
e
ex
p
e
r
im
en
t is r
ep
r
esen
ted
in
Fig
u
r
e
7
an
d
p
r
ed
icted
C
T
im
ag
e
f
r
o
m
MRI
a
s
s
h
o
wn
in
Fig
u
r
e
8
.
T
h
e
c
o
m
p
ar
is
s
io
n
r
esu
lt
o
f
SIFT
an
d
SUR
F a
s
s
h
o
wn
in
T
ab
le
2
.
Fig
u
r
e
7
.
(
a
)
C
T
I
m
ag
e
,
(
b
)
M
R
I
I
m
ag
e,
(
c)
C
T
C
alcu
latio
n
,
(
d
)
SUR
F f
ea
tu
r
es,
(
e
)
&
(
h
)
Hig
h
Dim
en
s
io
n
al
Sp
ac
e,
(
f
)
&
(
i)
C
o
r
r
esp
o
n
d
in
g
L
NDs,
(
g
)
&
(
j)
SDL
an
d
(
k
)
KNN
R
eg
r
ess
io
n
Fig
u
r
e
8
.
Pre
d
icte
d
p
C
T
im
ag
er
y
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
N
o
va
l a
d
va
n
ce
n
o
n
-
lin
ea
r
d
es
crip
to
r
a
n
d
ch
a
r
a
cteris
tic
eq
u
iva
len
cy
to
…
(
S
o
w
mya
B
a
ch
u
)
55
T
ab
le
2
.
C
o
m
p
a
r
is
o
n
o
f
SIFT
an
d
SUR
F b
y
Usi
n
g
PS
NR
an
d
MA
E
M
e
t
h
o
d
s/
P
a
r
a
me
t
e
r
s
P
S
N
R
M
A
E
S
I
F
T
2
7
.
5
0
2
1
8
0
.
9
5
7
9
S
U
R
F
3
4
.
3
7
7
6
8
9
.
8
0
3
6
5.
CO
NCLU
SI
O
N
D
u
r
in
g
th
is
ass
ess
m
en
t,
we
s
u
g
g
est
a
co
n
s
titu
en
t
co
o
r
d
in
a
te
m
eth
o
d
b
y
m
ea
n
s
o
f
lear
n
ed
n
ea
r
b
y
d
escr
ip
to
r
s
u
s
ed
f
o
r
a
n
ticip
atin
g
C
T
o
n
o
r
af
ter
MR
p
ictu
r
e
in
f
o
r
m
atio
n
.
T
h
e
n
ec
ess
ar
y
d
escr
ip
to
r
s
o
f
th
e
m
ag
n
etic
r
eso
n
an
ce
p
ictu
r
e
a
r
e
p
r
im
ar
y
an
ticip
ated
to
war
d
a
h
ig
h
d
im
en
s
io
n
al
g
a
p
to
war
d
g
e
t
n
o
n
-
lin
ea
r
d
escr
ip
to
r
s
u
tili
zin
g
a
n
o
p
en
c
o
m
p
o
n
en
t
d
r
awin
g
.
T
h
ese
d
es
cr
ip
to
r
s
ar
e
h
ig
h
e
r
th
r
o
u
g
h
th
e
im
p
lem
en
tatio
n
of
a
b
etter
SDL
co
m
p
u
tatio
n
.
T
h
e
ex
p
e
r
im
en
t
r
esu
lts
illu
s
tr
ate
to
th
e
lear
n
ed
n
o
n
lin
e
ar
d
escr
ip
to
r
s
a
r
e
s
u
cc
ess
f
u
l
u
s
ed
f
o
r
th
ic
k
c
o
o
r
d
in
a
tin
g
alo
n
g
with
p
C
T
ex
p
ec
tatio
n
.
I
n
a
d
d
itio
n
,
th
e
p
l
an
n
ed
C
T
esti
m
ate
ap
p
r
o
ac
h
ca
n
ac
c
o
m
p
lis
h
f
o
c
u
s
ed
ex
ec
u
tio
n
c
o
n
tr
asted
alo
n
g
with
a
f
ew
cu
ttin
g
-
ed
g
e
tech
n
iq
u
es.
F
UT
UR
E
WO
RK
D
u
r
in
g
o
u
tlo
o
k
s
tu
d
y
,
o
u
r
e
f
f
o
r
t
r
eso
lv
es
liv
e
ex
ten
s
iv
e
t
o
war
d
d
ee
p
lea
r
n
in
g
,
also
m
ed
ical
p
ictu
r
e
r
etr
iev
al
r
eso
lv
e
b
e
f
u
r
th
e
r
th
r
o
u
g
h
co
llect
a
lo
t
o
f
s
u
b
jects
u
s
ed
in
f
av
o
r
o
f
m
e
d
ical
ap
p
li
ca
tio
n
.
T
h
e
m
ed
ical
p
u
r
p
o
s
e
d
esig
n
ed
in
f
av
o
r
o
f
MRI
-
b
ased
cu
r
r
en
t RT
co
n
d
u
c
t p
r
ep
ar
atio
n
r
eso
lv
e
also
i
n
v
e
s
tig
ates
ca
r
ef
u
lly
.
RE
F
E
R
E
NC
E
S
[1
]
Zaid
i,
M
.
-
L.
M
o
n
tan
d
o
n
,
a
n
d
D.
O.
S
l
o
sm
a
n
,
"
Attrac
ti
v
e
re
v
e
rb
e
ra
ti
o
n
ima
g
in
g
-
g
u
id
e
d
les
se
n
i
n
g
a
n
d
d
isp
e
rse
re
m
e
d
ies
in
th
re
e
-
d
ime
n
sio
n
a
l
m
in
d
p
o
sitro
n
o
u
tfl
o
w
to
m
o
g
ra
p
h
y
,
"
M
e
d
ica
l
P
h
y
sic
s
,
v
o
l.
3
0
,
n
o
.
5
,
p
p
.
9
3
7
-
9
4
8
,
2
0
0
3
.
[2
]
S.
-
H.
Hs
u
,
Y.
Ca
o
,
K.
H
u
a
n
g
,
M
.
F
e
n
g
,
a
n
d
J.
M
.
Ba
lt
e
r,
"
E
x
a
m
in
a
ti
o
n
o
f
a
stra
teg
y
fo
r
c
re
a
ti
n
g
e
n
g
in
e
e
re
d
C
T
m
o
d
e
ls f
ro
m
M
RI
o
u
t
p
u
ts
o
f
t
h
e
h
e
a
d
a
n
d
n
e
c
k
fo
r
ra
d
iatio
n
trea
t
m
e
n
t,
"
Ph
y
sic
s in
p
re
sc
rip
ti
o
n
a
n
d
sc
ien
c
e
,
v
o
l
.
5
8
,
n
o
.
2
3
,
p
p
.
8
4
1
9
-
8
4
3
5
,
2
0
1
3
.
[3
]
D.
Iz
q
u
ierd
o
g
a
rc
ia,
A.
E.
Ha
n
se
n
,
S
.
F
ö
rste
r,
D.
Be
n
o
it
,
S
.
S
c
h
a
c
h
o
ff
,
S
.
F
ü
rst
,
K.
T.
Ch
e
n
,
D.
B.
Co
n
d
e
,
a
n
d
C.
Ca
tan
a
,
"
An
S
P
M
8
-
b
a
se
d
Ad
v
a
n
c
e
fo
r
Atten
u
a
ti
o
n
Co
rre
c
ti
o
n
Co
m
b
in
i
n
g
S
e
g
m
e
n
tatio
n
a
n
d
No
n
-
u
n
b
e
n
d
i
n
g
Tem
p
late
F
o
rm
a
ti
o
n
:
Ap
p
li
c
a
ti
o
n
to
S
imu
lt
a
n
e
o
u
s
P
ET
/
M
R
Bra
in
Im
a
g
in
g
,
"
J
o
u
r
n
a
l
o
f
N
u
c
lea
r
M
e
d
icin
e
,
v
o
l.
5
5
,
n
o
.
1
1
,
p
p
.
1
8
2
5
-
3
0
,
2
0
1
4
.
[4
]
M
.
Ho
fm
a
n
n
,
F
.
S
tei
n
k
e
,
V.
S
c
h
e
e
l,
G
.
Ch
a
rp
iat,
J.
F
a
rq
u
h
a
r,
P
.
A
sc
h
o
ff,
M
.
Bra
d
y
,
B.
S
c
h
ö
lk
o
p
f,
a
n
d
B.
J.
P
ich
ler,
"X
-
ra
y
Ba
se
d
Atte
n
u
a
ti
o
n
Co
rre
c
ti
o
n
fo
r
P
ET
/
M
RI:
A
No
v
e
l
Ad
v
a
n
c
e
Co
m
b
i
n
i
n
g
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
a
n
d
Atlas
Re
g
istratio
n
,
"
J
o
u
rn
a
l
o
f
N
u
c
lea
r M
e
d
ici
n
e
,
v
o
l.
4
9
,
n
o
.
1
1
,
p
p
.
1
8
7
5
-
1
8
8
3
,
2
0
0
8
.
[5
]
N.
Bu
rg
o
s,
M
.
J.
Ca
rd
o
s
o
,
K.
Th
iele
m
a
n
s,
M
.
M
o
d
a
t,
S
.
P
e
d
e
m
o
n
te,
J.
Dic
k
so
n
,
A.
Ba
rn
e
s,
R.
Ah
m
e
d
,
C
.
J.
M
a
h
o
n
e
y
,
J.
M
.
S
c
h
o
t
t,
J.
S
.
D
u
n
c
a
n
,
D.
Atk
i
n
so
n
,
S
.
R.
Arri
d
g
e
,
B.
F
.
Hu
t
to
n
,
a
n
d
S
.
Ou
rse
li
n
,
"
C
o
n
strict
io
n
Co
rre
c
ti
o
n
S
y
n
th
e
sis
fo
r
Hy
b
ri
d
P
ET
-
M
R
S
c
a
n
n
e
rs:
Ap
p
li
c
a
ti
o
n
t
o
Bra
in
S
t
u
d
ies
,
"
I
EE
E
T
r
a
n
sa
c
ti
o
n
s
o
n
M
e
d
ica
l
Ima
g
i
n
g
,
v
o
l.
3
3
,
n
o
.
1
2
,
p
p
.
2
3
3
2
-
2
3
4
1
,
2
0
1
4
.
[6
]
I.
M
é
ri
d
a
,
N.
Co
ste
s,
R.
A.
He
c
k
m
a
n
n
,
A.
Drz
e
z
g
a
,
S
.
F
ö
rste
r,
a
n
d
A.
M
a
ll
e
ts,
"
As
se
ss
m
e
n
t
o
f
a
fe
w
m
u
lt
i
-
m
a
p
b
o
o
k
stra
teg
ies
fo
r
P
S
EUDO
-
CT
a
g
e
in
c
e
re
b
ru
m
M
RI
-
P
ET
we
a
k
e
n
in
g
re
c
ti
fica
ti
o
n
,
"
IEE
E
T
we
lf
th
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
si
u
m o
n
Bi
o
me
d
ica
l
Ima
g
i
n
g
,
p
p
.
1
4
3
1
-
1
4
3
4
,
2
0
1
5
.
[7
]
V.
Ke
e
re
m
a
n
,
Y.
F
iere
n
s,
T.
Bro
u
x
,
Y.
De
De
e
n
e
,
M
.
Lo
n
n
e
u
x
,
a
n
d
S
.
Va
n
d
e
n
b
e
rg
h
e
,
"
X
-
ra
y
Ba
se
d
Atten
u
a
ti
o
n
Co
rre
c
ti
o
n
fo
r
P
E
T/
M
RI
Us
in
g
Ultras
h
o
rt
Ech
o
Ti
m
e
S
e
q
u
e
n
c
e
s,"
J
o
u
rn
a
l
o
f
N
u
c
lea
r
M
e
d
ici
n
e
,
v
o
l.
5
1
,
n
o
.
5
,
p
p
.
8
1
2
-
8
1
8
,
2
0
1
0
.
[8
]
A.
Jo
h
a
n
ss
o
n
,
M
.
Ka
rlsso
n
,
a
n
d
T.
Ny
h
o
lm,
"
CT
s
u
b
stit
u
te
g
o
t
fr
o
m
M
RI
g
ro
u
p
in
g
s
with
u
lt
ra
-
sh
o
rt
re
v
e
rb
e
ra
ti
o
n
ti
m
e
,
"
M
e
d
ic
a
l
P
h
y
sic
s
,
v
o
l.
3
8
,
n
o
.
5
,
p
p
.
2
7
0
8
-
2
7
1
4
,
2
0
1
1
.
[9
]
M
.
E
.
Je
n
s,
M
.
K.
Ha
n
s,
L
.
K
o
e
n
Va
n
,
H
.
H.
Ra
sm
u
s,
A.
L.
A.
J
o
n
,
a
n
d
A.
Da
n
iel
,
"
A
v
o
x
e
l
-
b
a
se
d
e
x
a
m
in
a
ti
o
n
f
o
r
M
RI
-
ju
st
ra
d
i
o
th
e
ra
p
y
o
f
t
h
e
m
in
d
u
ti
li
z
i
n
g
u
lt
ra
sh
o
rt
re
v
e
rb
e
r
a
ti
o
n
ti
m
e
s,"
P
h
y
sic
s
in
M
e
d
ici
n
e
a
n
d
Bi
o
l
o
g
y
,
v
o
l.
5
9
,
n
o
.
2
3
,
p
p
.
7
5
0
1
-
7
5
1
9
,
2
0
1
4
.
[1
0
]
S
.
Ro
y
,
W.
-
T.
Wan
g
,
A.
Ca
ra
ss
,
J.
L
.
R
u
ler,
J.
A.
B
u
tma
n
,
a
n
d
D.
L.
P
h
a
m
,
"
P
ET
Atten
u
a
ti
o
n
C
o
rre
c
ti
o
n
Us
in
g
S
y
n
t
h
e
ti
c
CT
fro
m
Ultras
h
o
rt
Ech
o
-
Ti
m
e
M
R
Im
a
g
in
g
,
"
J
o
u
rn
a
l
o
f
N
u
c
lea
r
M
e
d
ici
n
e
,
v
o
l.
5
5
,
n
o
.
1
2
,
p
p
.
2
0
7
1
-
2
0
7
7
,
2
0
1
4
.
[1
1
]
M
.
R.
Ju
t
tu
k
o
n
d
a
,
B.
G
.
M
e
rse
re
a
u
,
Y.
Ch
e
n
,
Y.
S
u
,
B.
G
.
R
u
b
i
n
,
T.
L.
S
.
Be
n
z
i
n
g
e
r,
D.
S
.
Lalu
s
h
,
a
n
d
H.
A,
"
M
R
-
b
a
se
d
we
a
k
e
n
in
g
a
m
e
n
d
m
e
n
t
f
o
r
P
ET
/M
RI
n
e
u
r
o
lo
g
ica
l
e
x
a
m
in
a
ti
o
n
s
wit
h
c
o
n
siste
n
t
e
ste
e
m
e
d
c
o
n
strictio
n
c
o
e
fficie
n
ts f
o
r
b
o
n
e
th
r
o
u
g
h
a
c
h
a
n
g
e
fro
m
R
2
*
to
CT
-
H
o
u
n
sfie
ld
u
n
it
s,
"
Ne
u
ro
Ima
g
e
,
v
o
l.
1
1
2
,
p
p
.
1
6
0
-
1
6
8
,
2
0
1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.