TELKOM
NIKA
, Vol. 11, No. 4, April 2013, pp. 1868
~18
7
0
ISSN: 2302-4
046
1868
Re
cei
v
ed
Jan
uary 2, 2013;
Re
vised Feb
r
uary
12, 201
3
;
Accepte
d
Febru
a
ry 23, 2
013
Speckle Reduction of Ultrasound Elastography with
Bilateral Filter
Xiaoming Zhou*
1
, Wen Liu
1
, Dong.C. Liu
2
1
Computer C
o
ll
ege of Sich
ua
n Univ
ersit
y
2
School of Co
mputer Scie
nc
e, Sichua
n Un
i
v
ersit
y
Che
ngd
u 61
00
65, Chi
na,
+
8
6
-
28-8
512
52
69-
100
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: softzhou6
21
@16
3
.com
A
b
st
r
a
ct
Ultraso
und
el
a
s
tograp
hy h
a
s
bee
n w
e
l
l
a
p
p
lie
d i
n
early
tumor
di
ag
nosi
s
for obta
i
ni
ng
tissu
e
stiffness infor
m
ati
on. E
l
asto
grapyy
ma
y
pr
ovid
e us
eful
cli
n
ical
i
n
for
m
ati
on for
the
tiss
ue c
haract
e
ri
zation
.
But ultras
on
ic
w
a
ve interfer
e
n
ce w
i
l
l
pr
od
u
c
e sp
eckle
i
n
both
ph
ase
an
d e
n
vel
o
p
e
. S
o
i
n
co
nve
n
tio
n
a
l
ultraso
und
ela
s
tograp
hy, the
r
e are no
ise a
r
tifacts
w
h
ich prod
uce so
me
misd
ia
gnos
is. In this paper,
w
e
i
n
ve
stig
a
t
e b
i
l
a
te
ra
l
fil
t
e
r
de
-no
i
si
ng
me
th
od
to
re
du
ce
the
speckl
e. Bec
a
use th
e
bil
a
ter
a
l filt
er d
e
-n
ois
i
ng
meth
od c
an
gr
eatly s
m
o
o
th t
he sp
eckl
e at
the sa
me
ti
me
protect the
le
sion
ed
ge w
e
l
l
,
it has b
e
e
n
w
e
ll
prove
d
go
od i
m
p
a
ct in B-
mo
de. But in u
l
tra
s
oun
d el
as
togr
aphy, the
bil
a
teral filt
er
hasn
’
t been us
ed. S
o
w
e
use the
bil
a
ter
a
l filter to re
d
u
ce artifacts to prov
e the p
e
rformanc
e of
this metho
d
. In the exp
e
ri
ment,
beca
u
se
of the
bilat
e
ral fi
lter
de-n
o
isi
ng met
hod,
the no
ise artifacts
w
ill be
reduc
ed l
a
rg
el
y. W
e
use SN
Re
and
CN
Re
to
verify the
p
e
rformanc
e of
th
e b
ilater
a
l
filte
r
an
d fi
nal
ly t
h
is
metho
d
pr
oved
a
sig
n
ific
an
t
improve
m
ent to SNRe a
nd C
NRe.
Ke
y
w
ords
: ultr
asou
nd e
l
astog
r
aphy, bi
latera
l
filter, SNRe, CNRe
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
In the past, the palp
a
tion
wa
s the only
primar
y te
ch
nique to e
s
timate the stiffness of
tissu
e espe
ci
ally brea
st. But palpation
has la
rge
un
certai
nty and
limitations in
diagno
si
s. The
ultrasoun
d el
astogta
phy h
a
ve prop
osed
by Ophir
et al [1]. in 1991 whi
c
h is a
new te
chni
qu
e to
provide d
o
cto
r
s mo
re
clinic information.
As we
kno
w
,
high quality
displa
cem
e
n
t
estimation i
s
a majo
r st
ep. So the iterative
pha
se zero m
e
thod is u
s
e
d
for displa
ce
ment estimati
on [2].However, there a
r
e
noise artifact
s in
the ultra
s
o
u
n
d
ela
s
tog
r
ap
hy. So a lot
of de-noi
sing
method
s
are u
s
ed to
re
duce the
arti
fact
inclu
d
ing
co
mpoun
ding, fi
lters a
nd wavelet tran
sfor
m. And the bi
lateral filter d
e
-noi
sin
g
met
hod
is sho
w
n to la
rgely u
s
e in i
m
age p
r
o
c
e
s
sing
and it ha
s be
en ap
plie
d in B-mo
de [
3
]. Becau
s
e t
he
bilateral
filter de
-noi
sing
met
hod
can
greatly
smo
o
th the
spe
ckle
at the
same
time p
r
ote
c
t
the
lesio
n
edge
well, we use the bilateral filter de
-noi
sing
method in thi
s
pap
er.
2. Rese
arch
Metho
d
The bilate
ral
filter is n
o
ise
smoothi
ng a
n
d
edg
e-p
r
e
s
e
r
ving filter. T
he inten
s
ity value of
pixel is re
pla
c
ed by a
wei
ghted ave
r
ag
e of intens
ity
values from
near
pixels a
nd the weigh
t
is
based on G
a
ussian di
strib
u
tion. The bil
a
teral filter fu
nction i
s
defin
ed as:
1
[]
(
)
(
)
Sr
pp
q
q
qS
p
B
FX
G
p
q
G
X
X
X
W
(1)
2
2
1
()
,
e
x
p
2
2
sr
pp
q
qS
x
WG
p
q
G
X
X
G
x
(2)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Speckle Red
u
ction of Ultraso
und Ela
s
togra
p
h
y
with
Billateral Filte
r
(Xiaom
ing Z
hou)
1869
Whe
r
e
S
is the
p-cent
ric set
of local pixel,
q
is th
e el
eme
n
t of . And
||
||
pq
is the g
eomet
ri
c
distan
ce b
e
twee
n
q
and
p
as the ce
nter point in the set.
Pq
X
X
is the light distan
ce
betwe
en
q
and
p
as the center
point in the set.
As we
kno
w
, the iterative pha
se
zero method i
s
a useful m
e
thod to e
s
timate the
displ
a
cement
. And the t
w
o
pre
c
o
ndition
s: on
e is
moti
ons are o
n
ly ultrasoun
d
scan line
di
re
ction
and th
e
se
co
nd i
s
rigid
m
o
tion exi
s
ts [4]. But in fa
ct, it is im
po
ssible. In
mo
d
i
fied ph
ase-zero
method, we p
e
rform t
w
o-di
mensi
onal
cross co
rrelatio
n cal
c
ulatio
n whi
c
h is
sho
w
n bel
ow [5]:
,1
,1
,1
,0
,
,
1
/2
/2
0
*
,1
0
/2
/2
0,
;
1
ar
g
,
2
,
xy
xy
xy
xx
y
x
y
Tw
n
jw
xy
nT
w
e
y
l
x
t
y
x
t
dt
dx
(3)
Whe
r
e th
e va
riable
,
x
y
means the di
spla
ce
ment of the
th
x
scan at th
e
th
y
position alon
g scan
line di
re
ction.
The
numb
e
r
n
is the
ho
rizo
ntal win
d
o
w
size, an
d
Tw
stan
ds fo
r the
axial win
d
o
w
siz
e
.
V
a
ria
b
les
1
y
and
2
y
are
baseban
d complex si
gna
ls of the p
r
e
v
ious a
nd th
e cu
rrent
frame resp
ect
i
vely.
3. Results a
nd Analy
s
is
We use
the Model 049
Elasticity
QA p
hantom
,
CRI
S Virginia,
USA, to obtain
our pre-
and p
o
st-co
m
pre
ssi
on
RF
data from
a commercial
ultrasoun
d sy
stem scan
ne
r, the Saset iMa
go
c21.And
the
RF d
a
ta con
s
ists of
512
sa
mples with
4
0
MHz
sampli
ng fre
quen
cy
and the
cent
ral
freque
ncy a
s
7.5MHz.
Experiment h
a
s three p
r
im
ary step
s: firstly usin
g mo
dified ph
ase
-
zero meth
od
on I/Q
s
i
gnal to
es
timate the delay and
linear interpolation is us
ed in
co
rrelation
calc
ulation with s
h
ift.
The
sampl
e
wind
ow fo
r d
i
spla
cem
ent
cal
c
ulatio
n is 2mm
with 7
5
% overla
p
and the
itera
t
ion
numbe
r is 1;
the se
con
d
st
ep is filtere
d
with
a seve
n point Savitzky-Golay digit
a
l differentiati
on
filter to get the elasto
gra
m
s of the tissu
e and the
fin
a
l step is a
p
p
l
ying 7*7 wid
o
ws bilateral filter
to s
t
rain image.
To test
pe
rformance of th
e
bilateral filteri
ng meth
od,
we
use SNRe to te
st the
spe
c
kle
redu
ction:
e
SNR
, where
and
are the lumin
ance mean a
nd the stan
d
a
rd deviatio
n
resp
ectively
of the luminance in
side a
19*19
windo
w excludi
ng
the stru
cture and wi
re
targ
ets. We use 3
wind
ows to compute the S
NR a
nd e
s
timate the mean
from 3 group
s of SNRe.
We u
s
e the CNRe to estim
a
tion the co
ntrast en
han
ce
ment in strai
n
image [6]:
2
22
2(
)
sb
s
t
e
sb
st
CNR
, where
s
t
and
s
b
denote
the
mean
strain
of the ta
rg
et and
the
backg
rou
nd,
s
t
and
s
b
rep
r
e
s
e
n
t the standa
rd deviation
of
the strain in the target and the
backg
rou
nd resp
ectively.
Figure 1(a
)
. Conve
n
tional
Strain Image
Figure 1(b
)
. T
he Strain Ima
ge used the
Bilateral Filter
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No. 4, April 2013 : 1868 – 1
870
1870
In Figure1, we found th
at the strai
n
im
age
with
bila
teral filter i
s
useful to
red
u
ce th
e
spe
c
kle. And
we
we fou
n
d
that the SNRe a
nd
C
NRe with the
bil
a
teral filter
d
e
-noi
sin
g
met
hod
are
39.71
96
and
3.218; t
he SNRe
an
d CNRe
of
convention
a
l strain im
age
are
17.77
2 a
n
d
2.4758. S
o
S
N
Re a
nd
CNRe
with
the
bilateral
filter de
-noi
sing
method
have
bee
n im
prov
ed
222.8% and
129.9% re
sp
ectively. So the bilateral f
ilter de-noi
sin
g
method i
s
useful in
stra
in
elasto
gra
phy.
4. Conclusio
n
In this
pap
er,
we
p
r
op
ose
d
bilate
ral
filter
de-noi
sing
metho
d
to
redu
ce th
e a
r
t
i
facts i
n
strain
ima
ge
and
prese
r
ve
d the
edg
e
of lesi
on i
n
cl
usi
on. Firstly, we ap
ply the
modified
pha
se-
zero metho
d
to get the strain image a
n
d
the se
con
d
step is the bil
a
teral filter de
-noi
sing meth
od
to redu
ce the
speckle. In the experim
e
n
t, with
the bilateral filter de-n
o
isi
ng m
e
thod the noi
se
artifact
s h
a
s bee
n
red
u
ced la
rgely
a
nd a
ppli
c
atio
n of thi
s
m
e
thod
prove
d
a
sig
n
ifica
n
t
improvem
ent to SNRe an
d CNRe.
Referen
ces
[1]
Ophir J,
Ce
spede
s I, Pon
nekanti H, Y
a
zdi Y,
Li X.
Elastog
r
ap
hy: A Quantitati
v
e Method fo
r
Imaging the
Elasticity of Biologi
cal Tiss
ues. Ult
r
ason
Imaging. 199
1; 13(2
)
: 111-134.
[2]
Andre
a
s Pe
savento, Ch
ristian Perrey, Martin
Kr
ue
ge
r
m
, H
e
lmu
t
Er
me
rt. A time-effic
i
ent and
acc
u
rate
s
t
rain es
timation c
o
n
c
e
p
t for ultra
s
oni
c el
astog
r
ap
hy u
s
ing ite
r
ative
pha
se
ze
ro
estimation.
IEEE Trans
. Ult
r
ason, Fe
rrocelec
t
Freq. Contr
. 1999; 4
6
(5
).
[3]
S Paris, P Ko
rnp
r
ob
st, JT
Tumblin. Bilat
e
ral
Filte
r
ing:
Theo
ry and
Applicatio
ns.
Found
ation
s
and tren
ds in
comp
uter g
r
a
phics an
d visi
on [1572
-27
4
0
] Paris. 200
8; 4:1.
[4]
Andre
a
s Pesavento, Chri
stian Perrey,
Martin
K
r
ue
g
e
r, Helmut E
r
mert. A Tim
e
-Efficient a
n
d
Ac
c
u
rate Strain Es
timation Co
n
c
ept fo
r
Ultra
s
oni
c
Elastog
r
ap
hy Using
Iterati
ve Phase Z
e
ro
Estimation. IEEE Trans. on ultrasoni
cs,
ferr
oel
ectri
cs,and frequency c
ontrol. 1995; 46(5).
[5]
Zhiqian
g
Jia
n
g
, Paul Liu, Dong
C, Liu.
Modified Pha
s
e
Zero Meth
od
for Ultra
s
ou
n
d
Free
han
d
Strain Im
aging
. Pro
c
. 3
r
d Intern
ation
a
l Co
nferen
ce on Bioi
nfo
r
matics
and
Biomedi
cal
Enginee
ring. 2009;
1-4.
[6]
Bilgen M, In
sana MF. P
r
e
d
icting ta
rg
et detecta
bility in acou
stic
el
astog
r
ap
hy.
IEEE Ultrason
Sym
pos.
200
7; 2: 1427-1
4
30.
Evaluation Warning : The document was created with Spire.PDF for Python.