Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
1
,
Jan
uar
y
201
9
,
pp.
205
~
2
09
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
1
.pp
205
-
20
9
205
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Extracti
ng acous
tic shado
wing fr
om
ultras
ound
image usin
g
local diff
erence
Hy
u
n
Ju
n
Par
k
1
, K
w
ang
Ba
ek Kim
2
1
Division
of
Soft
ware
Conve
rge
n
ce
,
Cheong
ju
Un
ive
rsit
y
,
Cheong
ju
28503,
Korea
2
Depa
rtment of
Com
pute
r
Engi
n
ee
ring
,
Sil
la Uni
ver
sit
y
,
Busan
4
6958,
Kore
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Sep
1
0
, 201
8
Re
vised N
ov 11
, 2
018
Accepte
d Nov
19
, 201
8
W
e
propose
a
m
et
hod
for
ext
r
ac
t
ing
ac
ousti
c
shadowing
from
ult
rasound
image.
If
we
lo
c
at
e
th
e
acous
ti
c
shadowing
exa
c
t
l
y
the
n
we
c
an
a
lso
ext
ract
the
h
y
per
ec
hoi
c
subs
ta
nce
s
wit
h
a
high
proba
b
il
ity
.
Th
e
proposed
m
et
hod
rec
onsti
tut
es
an
origi
nal
ul
tra
sou
nd
image
to
simpl
if
y
calc
ul
at
ion
s
and
uses
the
lo
ca
l
l
y
hori
z
ont
al
d
iffe
r
ence
to
ext
r
act
shado
w
ca
ndid
at
es.
T
he
shadow
ca
ndid
at
es
ar
e
c
la
ss
ifi
ed
in
to
sta
rt
and
fini
sh
poi
nt,
and
shadow
reg
ions
are
ext
ra
ct
ed
b
y
usi
ng
the
m
.
The
ex
per
iment
result
s
show
the
proposed
m
et
hod
ext
ra
ct
s
23
of
27
ac
ousti
c
shado
ws
from
ult
rasonog
ram
eff
icientl
y
and
it
can
be
used
in
wid
e appl
i
ca
t
ions.
Ke
yw
or
ds:
Acousti
c s
hado
wing
Gall
blad
der
Gall
ston
e
Sh
a
dow
e
xtrac
ti
on
Ultraso
nogram
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Kw
a
ng Bae
k Kim
,
Dep
a
rtm
ent o
f C
om
pu
te
r
E
ng
i
neer
i
ng,
Sil
la
U
niv
e
rsit
y,
Busan 4
6958,
Korea
.
Em
a
il
:
gb
kim
@sil
la
.ac.kr
1.
INTROD
U
CTION
Ultraso
nogra
phy
is
of
te
n
the
fir
st
i
m
aging
te
st
done
in
pe
op
le
w
ho
ha
ve
sym
pto
m
s
fr
om
abdom
en
reg
i
on.
It
is
an
easy
te
st
to
ha
ve
done
,
an
d
it
use
s
no
ra
diati
on
[1
]
.
Also
ultra
so
no
gr
a
phy
is
us
ef
ul
for
veteri
nar
y
m
edici
ne.
It
us
es
no
an
est
hetic
,
and
it
is
no
t
expensive
to
diagnosis
anim
als
[2
]
.
Fo
r
a
gallbla
dd
er
ultraso
und
e
xam
,
you
sim
ply
l
ie
on
a
ta
bl
e
w
hile
the
do
ct
or
m
ov
es
the
trans
ducer
al
ong
the
sk
in
over
the
ri
gh
t
uppe
r
a
bdom
en.
Ob
se
r
ved
s
hape
an
d
siz
e
of
ti
ssu
e
from
ultrasono
gr
am
are
us
e
d
f
or
di
agnosin
g.
Als
o,
a
co
us
ti
c
sh
a
dowing
is
al
so
us
e
d.
The
acou
sti
c
sh
a
dowi
ng
us
ed
to
cause
by
ab
norm
al
su
bs
ta
nc
es.
That
m
ea
ns
the
sh
a
dows
i
n ult
r
aso
nogr
am
r
ep
resen
t
diseases
o
ccasi
on
al
ly
.
Ca
lc
ific
at
ion
,
ston
e
,
bo
ne,
a
nd
ai
r
a
re
ty
pi
cal
hyper
ec
ho
ic
su
bst
ances
that
occ
ur
pos
t
acou
sti
c
sh
a
dowing
in
ultraso
nogram
.
Stat
ed
rev
e
rse
ly
,
if
we
locat
e
the
acous
ti
c
sh
a
dowing
the
n
we
ca
n
al
so
extract
the
hype
rechoi
c
su
bs
ta
nces.
So
m
e
hyper
ec
ho
ic
s
ub
sta
nce
s
su
ch
as
cal
c
ific
at
ion
an
d
s
ton
e
are
of
te
n
fatal
,
so
it
m
us
t be
f
ound
durin
g ul
traso
nograp
hy
[3,
4].
The
ac
ou
sti
c
s
hado
wing
ha
s
been
us
i
ng
as
an
im
po
rtant
f
act
or
in
pr
a
ct
ic
al
diagnosis.
T
her
e
a
re
the
fiel
d
of
m
edicin
e
us
i
ng
ac
ous
ti
c
sh
adowin
g
extracti
on
su
c
h
as
gallst
one
extracti
on,
ki
dney
ston
e
pr
e
di
ct
or
,
cal
ci
fied
plaq
ue
detect
ion
,
a
nd
so
on
[5
-
7].
Ther
e
f
or
e,
if
w
e
can
locat
e
th
e
acou
sti
c
sh
a
dowi
ng
a
uto
m
atical
ly
,
it
w
il
l help t
he practi
cal
d
ia
gnos
is i
n va
rio
us
fiel
ds
of m
edic
al
.
Un
ti
l
recently
,
there
ha
s
bee
n
a
la
ck
of
re
sea
rch
on
a
uto
m
atical
ly
extracti
ng
the
aco
us
ti
c
sh
a
dowing
by
us
in
g
m
edical
i
m
age
processin
g
m
et
ho
ds.
The
re
ar
e
so
m
e
research
es
prese
nt
a
sh
ad
ow
extr
act
ing
al
gorithm
,
bu
t
their
al
gorithm
s
are
li
m
i
te
d
f
or
sp
eci
fic
ap
pl
ic
at
ion
.
T
he
r
esearche
s
on
e
xtracti
ng
the
a
coust
ic
sh
a
dowing
f
r
om
ultraso
nogra
m
can
be
us
e
d
in
wide
a
ppli
cat
ion
s
are
sti
ll
need
e
d.
T
he
refor
e
,
in
t
his
pap
e
r
,
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
2
0
5
–
2
0
9
206
we
propose
a
m
et
ho
d
for
e
x
t
racti
ng
aco
us
ti
c
sh
a
dowing
from
ultrasoun
d
i
m
age
by
usi
ng
local
dif
fer
e
nce
to
us
e i
n wide a
ppli
cat
ion
s.
This
pap
e
r
is
orga
nized
as
f
ollows.
Sect
i
on
2
e
xp
la
ins
a
pr
e
processi
ng
m
et
ho
d
a
nd
s
ha
dow
reg
i
on
extracti
on
m
eth
ods.
Sect
io
n
3
e
valuates
t
he
perf
or
m
ance
of
the
pro
posed
m
et
ho
d
us
i
ng
twenty
im
ages
wh
ic
h
wer
e
u
se
d
i
n
a
ct
ual in
-
hosp
it
al
d
ia
gnose
s. F
inall
y, Sect
ion
4 pr
ese
nts t
he c
on
cl
us
io
ns
.
2.
E
X
T
R
AC
TI
N
G ACO
US
TI
C
S
HADO
WI
NG
Figure
1
s
hows
overall
proces
s for
extracti
ng
sh
a
dow re
gion
s.
Figure
1. O
veral
l
process
for sha
dow
e
xtract
ion
2
.
1.
Pre
proce
ssing
Be
cause
ultras
onogra
phy
us
e
s
sound
wa
ve
s
to
gen
e
rate
i
m
ages
of
inte
rn
al
orga
ns
,
t
he
ac
ou
sti
c
sh
a
dowing
has
directi
onal
it
y.
To
extr
act
the
sh
a
dow
f
ro
m
the
ori
gi
nal
im
age,
it
has
t
o
r
epeate
dly
us
e
t
he
si
n
a
nd
c
os
operati
on
s
for
cal
cula
ti
ng
coor
din
at
e
s.
It is o
ne
cau
se o
f
inc
reasin
g
the pro
ces
sin
g
tim
e. Th
eref
or
e we
reconsti
tute t
he
i
m
age to
sim
plify ca
lc
ulati
ons.
Figure
2
s
h
ow
s
an
e
xam
ple
of
im
age
reconsti
tuti
on.
At
f
irst,
as
s
how
n
in
Fig
ur
e
2
(a),
we
fin
d
the
locat
ion
of
P
1
to
P
4
a
nd
cal
culat
e
the
van
i
sh
in
g
po
i
nt.
A
fter
the
n,
we
r
econst
it
ute
the
i
m
age
as
sho
wn
in
Figure
2
(b)
usi
ng
piece
wise
aff
i
ne
warpin
g
[8
]
.
N
ow,
w
e
can
j
us
t
c
ons
ider
ver
ti
cal
directi
on
to
e
xtr
act
the
sh
a
dowing i
ns
te
ad of
sin
a
nd
cos o
per
at
io
ns
.
Figure
1
.
I
m
age r
ec
on
sti
tuti
on
for
e
xtracti
ng s
ha
dow
In
reconsti
tute
d
im
age,
gen
e
rall
y,
the
aco
ust
ic
sh
ad
ow
i
ng
ha
s
a
ve
rtic
al
ly
lon
g
a
nd
dark
s
hap
e
,
bu
t
le
vel
of
da
rk
is
not
an
abs
olu
te
value
.
Ther
e
f
or
e
,
w
e
hav
e
to
u
se
the
relat
ive
di
ff
ere
nce
to
e
xtract
the s
hado
w.
As
a
pr
e
proces
sing,
we
a
pply
Gaussi
an
filt
er
to
rem
ov
e
noise
an
d
m
edian
filt
er
to
e
xt
ract
sh
a
dow
cand
i
dates.
3×
15 size
of
Ga
ussi
an fil
te
r
is u
sed. It als
o
m
a
gn
i
fies the
hori
zon
ta
l
differe
nc
e.
2
.
2
.
E
xt
r
ac
ti
ng
th
e
Sh
ado
w
R
egi
on
s
Figure 3
sho
ws
resu
lt
s o
f
t
he
ver
ti
cal
an
d
horizo
ntal
diff
e
re
nce
cal
culat
io
n.
The
ve
rtic
al
diff
ere
nce
is
no
t
su
it
able
f
or
extracti
ng
t
he
sh
a
dow
as
s
hown
in
Fig
ur
e
3
(a),
s
o
we
us
e
the
horizo
ntal
diff
e
re
nce
s
ho
wn
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ext
ra
ct
ing
acoustic
sha
do
wi
ng fr
om
ultra
sound i
mage usi
ng loc
al
diff
eren
ce
(
Hyun
J
un
Park
)
207
Figure
3
(
b).
E
ven
if
pix
el
s
ne
ar
the
s
ha
dow
reg
i
ons
are
ra
pid
ly
cha
nged
,
ad
j
acent
pix
el
s
ha
ve
sim
i
la
r
value.
Ther
e
f
or
e
we
cal
culat
e
the
horizo
ntal
dif
f
eren
ce
by
us
i
ng
pi
xels
ix
a
nd
ix
-
5
(x
is
an
in
de
x
of
x
-
axis
).
The regi
ons
ha
ve big
ger h
or
iz
on
ta
l
diff
e
r
e
nc
e than t
hresh
ol
d
is ass
um
ed
to
the s
ha
dow
ca
nd
i
dates.
Figure
2
.
V
e
rtic
al
an
d h
ori
zo
ntal dif
fer
e
nce
Figure
4
s
how
s
sh
a
dowing
r
egio
n
e
xtracti
ng
proces
s.
Fi
gure
4
(a
)
s
hows
re
gion
of
interest
f
or
extracti
ng
the
acoust
ic
sh
a
dowing.
Fi
gure
4
(b)
s
hows
t
he
sh
a
dow
ca
nd
i
dates.
T
he
s
ha
dow
ca
nd
i
date
s
are
cl
assifi
ed
as
sha
dow
sta
rtin
g
and
fi
nish
i
ng
point.
We
de
fin
e
the
s
ha
dow
s
ta
rting
points
a
s
a
pix
el
wh
ic
h
is
it
s
ver
ti
cal
dif
fer
e
nce
is
ra
pid
ly
decr
ease
d.
I
n
c
on
t
rast,
the
fin
ishin
g
points
a
re
de
fine
d
as
a
pix
el
that
it
s
ve
rtic
al
diff
e
re
nce
is
rap
i
dly
increa
sed.
Fig
ur
e
4
(c)
s
hows
fi
nally
resu
lt
s
of
extracti
ng
the
s
hado
w
r
egio
ns.
The
sh
a
dow
ca
n
be
ext
racted
by co
nnect
in
g
the sh
a
dow
sta
rting
po
i
nts and f
inis
hing points.
Be
f
or
e ext
racti
ng
the s
hado
w
re
gi
on
s
, th
e
v
e
rtic
al
ly
s
m
all size of sha
dow
can
did
at
es a
re
rem
ov
e
d by
us
in
g si
ze fil
te
r.
Figure
3
.
Proce
ss of e
xtracti
ng
sh
a
dow re
gion
s
3.
E
X
PERI
MEN
TAL RES
UL
TS
The
s
oft
ware
was
im
ple
m
ented
in
C+
+,
a
nd
the
Mi
cro
s
of
t
Visu
al
20
10
In
te
gr
at
e
d
D
evelo
pm
ent
En
vironm
ent
was
us
e
d
in
t
he
co
ding.
Sim
u
la
ti
on
s
w
ere
run
on
an
IBM
-
c
om
patible
PC
with
a
n
I
ntel
P
entium
i7
-
2640M C
P
U @
2.80GH
z
with
8.0G
B
RAM.
Twe
nty
ultras
ound
im
ages
wer
e
us
e
d
f
or
our
ex
pe
rim
e
nt.
T
here
are
total
27
s
hado
ws,
an
d
t
he
exp
e
rim
ental
r
esults are s
how
n
in Fi
gure
5
a
nd
Table
1.
In
Figure
5
an
d 6, do
tt
ed
li
nes
in
dicat
e act
ual sh
ad
ow
reg
i
on
s
an
d sol
id li
nes
a
re e
xt
racted
reg
i
ons
by the
pro
pose
d
m
et
ho
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
2
0
5
–
2
0
9
208
Figure
4
.
Resul
ts of aco
us
ti
c s
hado
w
e
xtracti
ng
Table
1
. E
val
ua
ti
on
of the
Propose
d
Me
th
od
Extractin
g
Resu
lt
Po
sitiv
e
Neg
ativ
e
Sh
ad
o
w (
Tr
u
e)
2
3
(
Hit)
2
(
Miss)
No
n
-
sh
ad
o
w (
False)
5
(
False al
ar
m
)
-
Ex
per
im
ental
resu
lt
s
show
s
23
of
27
s
ha
dows
are
e
xtract
ed,
5
sh
a
dow
s
are
false
al
ar
m
ed
and
2
sh
a
dows
a
re
m
issed.
We
t
ried
to
m
ini
m
iz
e
t
he
m
iss
cases
than
the
false
al
arm
to
util
iz
e
the
pro
po
se
d
m
et
hod
as a to
ol fo
r pra
ct
ic
al
d
ia
gnosi
s.
Figure
6
s
hows
e
xam
ple
of
fail
ed
extrac
ti
on
s.
I
n
4
an
d
18
im
ages,
sh
a
dows
ha
ve
not
en
ou
gh
horizo
ntal
di
fference
to
extra
ct
the
s
hado
ws
,
s
o
they
are
m
issed.
I
n
7
a
nd
16
im
ages,
extracte
d
re
gions
a
re
si
m
il
ar w
it
h
sha
dow, so
they
are f
al
se
alarm
ed.
Figure
6. Exa
m
ple o
f
fail
ed
extracti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ext
ra
ct
ing
acoustic
sha
do
wi
ng fr
om
ultra
sound i
mage usi
ng loc
al
diff
eren
ce
(
Hyun
J
un
Park
)
209
4.
CONCL
US
I
O
N
In
this
paper,
we
prese
nt
a
ne
w
m
et
ho
d
f
or
extracti
ng
the
acou
sti
c
s
hadow
i
ng
from
ult
rason
ogram
.
It
us
es
the
lo
cal
ly
ho
rizo
ntal
diff
ere
nce
to
extract
the
sh
a
dow
can
did
at
es,
an
d
ext
ract
the
sh
ad
ow
by
connecti
ng t
he sha
dow
sta
rtin
g po
i
nt and
fini
sh
in
g p
oin
t.
The
e
xperim
e
nt
res
ults
s
how
the
pr
opose
d
m
et
ho
d
ex
tract
s
23
of
27
ac
ousti
c
sha
dows
f
ro
m
ultraso
nogram
eff
ic
ie
ntly
.
Howe
ver,
ther
e
are
al
so
5
false
al
ar
m
s
and
2
m
isses.
For
f
ur
the
r
work
s
,
extracte
d
s
hadow
re
gions
by
the
propose
d
m
et
ho
d
can
be
im
pr
ov
e
d
by
ver
ific
at
io
n
pr
ocess.
It
c
an
c
orrect
th
e
fail
ed
cases
.
The
pr
opos
e
d
m
et
ho
d
ca
n
be
ext
racti
ng
the
hype
rec
hoic
su
bs
ta
nces
by
extracti
ng
the
aco
us
ti
c
sh
a
dows, so i
t
can
be uti
li
zed
in v
a
rio
us
ultra
so
nic
m
edical
i
m
age pro
ces
sing sy
ste
m
s.
REFERE
NCE
S
[1]
Park
HJ
,
Kim
KB,
Cha
EY.
A
new
appr
oach
to
loc
a
te
th
e
gal
lbladder
fro
m
ult
rasound
image
using
int
e
nsit
y
stret
ch
ing
and
S
OM
-
base
d
col
or
quant
i
za
t
ion. Cu
rre
nt
Med
ical
Im
agi
ng
R
evi
ews.
2015;
11(1):
63
-
68.
[2]
Penninc
k
D,
d
'
A
njou
MA
.
Atl
as
of
sm
al
l ani
m
al
ult
rasonogra
ph
y
.
John W
il
e
y
&
S
ons.
2013.
[3]
Park
HJ
,
Kim
K
B,
Cha
EY
.
Im
prove
d
Method
t
o
Loc
a
te
the
Ga
ll
bla
dd
er
in
Son
ogra
m
Based
on
Support
Vec
to
r
Mac
hine.
Journal
of
Me
di
cal Imaging
and
Hea
lth Inf
orm
ati
cs
.
2
016;
6(7):
1566
-
1569.
[4]
Le
e
FC
,
Dunm
ire
B,
Harp
er
JD
.
,
Cunit
z
BW
,
Pa
un
M,
Bai
l
e
y
M,
Sorensen
M
D.
Ultra
sound
ac
ou
stic
shadow
wid
t
h
is
an
ac
cur
a
te
p
red
ictor
of
kidney
stone
siz
e.
The
Journal
of
the
Ac
oustic
a
l
Soci
e
ty
of
Ame
ri
ca
.
2014;
135(4):
2267
-
2267.
[5]
Basij
M,
Ta
ki
A,
Yaz
dchi
M.
Aut
omatic
shadow
enhanc
eme
nt
in
int
ra
vasc
ular
ult
ras
ound
(
IVUS
)
imag
es
.
In
Biom
edi
c
al En
gine
er
ing
(ME
CBM
E),
2014
M
iddl
e
Ea
st
Confe
ren
ce on, IEEE
Februa
r
y
2014;
309
-
312.
[6]
Gao
Z,
Hau
WK,
Zha
ng
H,
Zha
ng
YT.
Aut
o
matic
Detect
ion
of
Calc
if
i
ed
Pl
aque
wit
h
Ac
o
ustic
Shadowin
g
.
In
The Internatio
nal
Conf
ere
n
ce
on
Hea
l
th
Infor
m
at
ic
s.
Spring
er Inte
rna
ti
ona
l
Pu
bli
shing,
Janua
r
y
2014;
197
-
199
.
[7]
Noble
JA
,
Bou
ker
roui
D.
Ultr
a
sound
image
se
gm
ent
at
ion:
a
surve
y
.
M
edical
Imaging,
IE
EE
Tr
ansacti
ons
on
.
2006;
25(8):
987
-
1010.
[8]
Matt
hews
I,
Bak
er
S.
Acti
v
e
app
ea
ran
ce
m
odel
s
rev
isit
ed.
Int
ernati
onal
Journal
of
Computer
Vi
s
ion
,
2004;
60(2)
:
135
-
164.
Evaluation Warning : The document was created with Spire.PDF for Python.