TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 14, No. 2, May 2015, pp. 329 ~ 33
4
DOI: 10.115
9
1
/telkomni
ka.
v
14i2.737
5
329
Re
cei
v
ed
Jan
uary 22, 201
5
;
Revi
sed
Ap
ril 16, 2015; Accepted Ma
y
1, 2015
Automatic Extraction of Diaphragm Motion and
Respiratory Pattern from Time-sequential Thoracic MRI
Windra S
w
a
s
tika*
1,2
, Yoshitada Ma
su
da
3
, Takashi
Ohnishi
4
, Hideaki Ha
neis
hi
4
1
Chib
a Univ
ers
i
t
y
, Gradu
ate Schoo
l of Engi
n
eeri
ng, Medic
a
l
S
y
stem Eng
i
n
eeri
ng,
1-33 Ya
yo
i-ch
o
,
Chiba, Jap
an,
263-8
5
2
2
2
Ma Chun
g Uni
v
ersit
y
, F
a
cult
y of Science a
n
d
T
e
chnolo
g
y
,
Villa P
uncak T
i
dar N-0
1
, Mala
ng, Indo
nesi
a
, 651
51
3
Chib
a Univ
ers
i
t
y
Hos
p
ita
l
, 1-8-1 Inoh
an
a,
Chib
a, Jap
an, 260-
085
6
4
Chib
a Univ
ers
i
t
y
, Ce
nter for F
r
ontier Med
i
c
a
l Eng
i
ne
eri
ng,
1-33 Ya
yo
i-ch
o
,
Chiba, Jap
an,
263-8
5
2
2
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
w
i
ndr
a.s
w
asti
ka@gra
du
ate.chib
a-u.jp
A
b
st
r
a
ct
T
horacic
time-
s
equ
enti
a
l MRI
can
be
us
ed t
o
ass
e
ss d
i
a
p
h
r
ag
m
moti
on
p
a
ttern w
i
tho
u
t
expos
in
g
radi
ation
to su
bject. Cl
inic
ia
n
s
may e
m
ploy
the
moti
o
n
to
eval
uate th
e s
e
vere
ness
of c
h
ron
i
c o
b
structiv
e
pul
monary
dis
ease (
C
OPD). T
h
is study pr
o
pose
d
a
nov
e
l
meth
od of
di
ap
hrag
m motion extraction meth
o
d
on ti
me-s
eq
ue
ntial t
horac
ic M
R
I in s
agitta
l p
l
ane. Otsu
’
s
thr
e
sho
l
d
an
d acti
ve co
ntour
alg
o
rith
m ar
e us
e
d
to
obtai
n dia
phr
a
g
m b
o
u
n
d
a
ry. An auto
m
at
ic d
i
ap
hrag
m
moti
on trackin
g
an
d extraction of
respirat
ory patter
n
are also
p
e
rfor
me
d bas
ed on
the di
aphr
ag
m bo
un
dary.
A total
of
120
0 fra
m
es
ti
me-
s
equ
enti
a
l M
R
I i
n
sagittal
pla
ne
w
a
s obtain
ed
for total of 15
subjects
(8
h
ealthy v
o
lu
nte
e
rs an
d 7 CO
PD pati
ents). T
h
e
prop
osed
met
hod
succ
essfu
lly
extracts d
i
aphr
ag
m
moti
on
an
d res
p
ir
atory p
a
tterns
for b
o
th
he
a
l
thy
volu
nteers a
n
d
COPD patient
s.
Ke
y
w
ords
:
ma
gn
etic reso
nanc
e i
m
ag
in
g
,
chronic obst
r
ucti
ve pu
l
m
on
ary dise
ase, respir
atory pattern
,
dia
phra
g
m
motion
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Diap
hra
g
m i
s
a do
me-sh
a
pe respirato
r
y orga
n lo
cat
ed bel
ow the
lung th
at se
parate
s
che
s
t from the abdo
men.
It controls the move
men
t
of the lungs and the b
r
eathing p
r
o
c
ess
(inhal
ation a
nd exhal
atio
n). The
moti
on of
the di
aphragm
ca
n be u
s
e
d
to evaluate t
he
sev
e
r
ene
s
s
of
ch
roni
c o
b
st
ru
ct
ive p
u
l
mona
ry di
se
ase
(COP
D) [1]. Hen
c
e,
in the
pa
st
few
decade
s dia
p
h
rag
m
motio
n
has b
een a
s
sesse
d
in several
studie
s
[1-6]. In 19
85, Diame
n
t et al.
[2] extracted
diaphragm
motion fro
m
ultrasono
grap
hy to evaluate dia
phra
g
m mot
i
on
abno
rmalitie
s. Gerscovich
et al. [3] and
Boussu
g
e
s
et al. [4] use
d
M-mo
de ult
r
asono
graphy
to
record di
aphragm m
o
tion i
n
two dim
e
nsions.
Desp
ite its portability, real
-time
examination and
no ioni
zation
radiatio
n, due
to the nature
of ul
traso
n
o
g
rap
h
y the imaging
re
sult
does
not rev
eal
tissu
e den
sity and potentia
lly creat
e
s
art
i
facts. The u
s
e of magnetic re
son
a
n
c
e imaging, which
provide
s
m
o
re cle
a
r
and
d
e
tailed ima
g
e
s
of
soft tissu
e
, has be
en
prop
osed in
[5-7]. Howeve
r,
none of them
use a
u
tomati
c extra
c
tion to extract
diap
hrag
m motion
and its re
spi
r
atory pattern.
In this study, we focu
sed
on automatic ex
tractio
n
of diaphragm
motion from
a time-
seq
uential
th
ora
c
ic M
R
I i
n
sagittal
plane. T
he
ex
tractio
n
wa
s
perfo
rmed
to
15
subje
c
ts (8
healthy volun
t
eers an
d 7
COPD patie
n
t
s). We the
n
com
pared t
he stati
s
tical
analysi
s
of the
diaph
rag
m
motion extracte
d from health
y volunteers
and COPD p
a
tients.
2. Subjects a
nd Metho
d
s
This se
ction descri
b
e
s
the
image
a
c
q
u
i
s
iti
on a
nd the
method
s of
automatic
dia
phra
g
m
motion extra
c
tion inclu
d
ing
respiratory pa
ttern extractio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 2, May 2015 : 329 – 334
330
2.1. Image Acquisition
The MR ima
g
e
s were a
c
q
u
i
red u
s
ing 1.5
T
INTERA ACHIVA nova-dual (Philip
s
Medical
Systems) wh
ole-b
ody sca
nner
with a
16ch SENSE TORSO XL
Coil. A 2D balan
ce
d FF
E
seq
uen
ce
wa
s u
s
e
d
. Th
e i
m
aging
pa
ra
meters a
r
e
a
s
follo
w. SENSE factor:
2.2, flip an
gle:
45
o
,
TR: 2.2m
s, T
E
: 0.9ms, FO
V: 384mm, in
-plan
e
resolut
i
on 25
6x256
pixels a
nd
1.5x1.5mm
2
, s
l
ic
e
thic
knes
s
:
7.5mm, s
lic
e gap=
6
.0mm, s
c
an time: 150ms
/frame.
Normal breat
hing wa
s in
structe
d
for all subj
ect
s
duri
ng the acq
u
isition pro
c
e
ss
and total
of 1200 fram
es in sagittal plane
were o
b
tained for
e
a
ch
subj
ect. Image a
c
qui
si
tion experim
e
n
t
wa
s co
ndu
cte
d
unde
r an a
pproval of Ethica
l
Review
Board of Chib
a University.
2.2. Diaphra
g
m Motion Extra
c
tion
In orde
r to g
e
t diaph
rag
m
motion, we f
i
rst
defin
e a
region
of interest (ROI)
of the MR
image by
croppin
g
the i
m
age th
at covers t
he di
aphragm
bo
unda
ry. Typically, diap
hragm
boun
dary is l
o
cate
d in middle of MRI i
n
sagittal pla
ne. To cove
r the whole a
r
ea of diaph
ra
gm
boun
dary, we
first define two pa
ram
e
ters, w and h to
rep
r
e
s
ent the
width and h
e
i
ght of ROI.
There are two main ste
p
s
to extract
diaph
rag
m
motion. The first step i
s
to obtain
diaph
rag
m
b
ound
ary for the first fram
e
only using
a
c
tive conto
u
r
algorith
m
. On
ce the first frame
of diaph
rag
m
boun
dary i
s
obtaine
d, the
next step i
s
to extract the
diaph
rag
m
b
ound
ary for t
he
sub
s
e
que
nt frame
s
base
d
on norm
a
lized cro
s
s correl
ation (NCC) valu
e. The compl
e
te
pro
c
e
s
ses of
diaph
rag
m
motion extracti
on are a
s
foll
ows.
We first p
e
rfo
r
m
clust
e
rin
g
-based im
age
thres
holdi
ng
usin
g Ot
su’s thre
shol
ding
method
[8]. After the thresholdin
g
pro
c
e
ss, the
diaph
r
agm
area
be
came
clea
rly disti
ngui
sha
b
le from
other o
r
gan
s.
A mask is created above
the diaph
ra
g
m
bound
ary as seed poi
nt in order to trace
the diap
hra
g
m
boun
da
ry usin
g a
c
tive conto
u
r al
go
rithm [9]. An optimizatio
n
of the diap
hragm
boun
dary
det
ection
can
al
so
be
optimi
z
ed u
s
in
g a
m
e
thod
propo
sed by
Alfiansyah [10]
or [1
1].
Note that this proc
ess
is
only
perfo
rme
d
for th
e first
frame
only. F
i
gure
1
sh
ows the
process of
obtainin
g
dia
phra
g
m bo
un
dary of the first frame.
Figure 1. Obtaining dia
p
h
r
agm bou
nda
ry for the first frame
To get diap
hragm bou
nda
ry for the sub
s
eq
uent
fram
es, we utili
ze
one col
u
mn
matrices
template defi
ned by T
x
wh
ere
x
=1..
w
. T
he elem
ent o
f
matrix
T
x
is obtaine
d fro
m
pixel value
s
of
the ROI
at co
lumn the
x
. T
herefo
r
e, the
size of mat
r
ix
T
x
is
1 x h
, where
h
is th
e heig
h
t of th
e
ROI. It is also necessa
ry
to gene
rate
a 2D
spatio
temporal of colum
n
x
(Fi
gure 2(a
)).
T
he
locatio
n
of di
aphragm b
o
u
ndary at colu
mn
x
for th
e subsequ
ent fra
m
es i
s
define
d
by the high
est
NCC value b
e
twee
n the matrix
T
x
and the 2D sp
atio temporal
of the subseque
nt frame
s
at
colum
n
x
. T
he p
r
o
c
e
s
s i
s
repe
ated f
o
r
x
=
1..w
.
We de
noted
th
e po
sition
of
the di
aph
ra
g
m
boun
dary at l
o
catio
n
x
as
f
x
(i)
wh
ere
i
repre
s
e
n
ts th
e
i
th frame.
The
f
x
(i)
sh
o
w
s peri
odi
c
p
eaks
and
vall
eys a
s
soci
ated wit
h
re
spi
r
atio
n cycle
s
.
Fi
gure 2 ill
ustrates ho
w to
dete
r
mine the
lo
ca
tion
of diaphragm
boun
dary.
S
e
t
RO
I
fo
r t
h
e
firs
t
fr
a
m
e
Ot
su
T
h
r
e
s
hol
d
A
c
tiv
e
co
u
n
t
o
r
a
l
g
o
r
ith
m
w
h
Y
X
Y
X
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Autom
a
tic Extraction of
Diaphragm
Motion and
Re
spi
r
atory Pattern
from
… (Win
d
ra Swasti
ka)
331
(a)
(b)
Figure 2. Determin
e the lo
cation of dia
p
h
rag
m
bou
nd
ary. (a) Elem
ent of matrix Tx is obtaine
d
from the pixel
values of RO
I at column x and t
he spati
o
temporal is
gene
rated at
the colum
n
x
from the su
bseque
nt frame
s
; (b) T
he det
ected
di
aph
ra
gm motion at colum
n
x (fx(i)), is
rep
r
e
s
ente
d
by the white line
2.3. Respira
t
or
y
Patterns
Extrac
tion
Re
spiratory
pattern
s a
r
e
automati
c
all
y
extr
acted
from dia
p
h
r
a
g
m motion
that is
previou
s
ly ob
tained. The
e
x
traction of
re
spirat
ory p
a
ttern
s is
only p
e
rform
ed at t
he column
x
t
hat
has the la
rge
s
t diaph
rag
m
movement.
(a)
(b)
(c
)
(d)
Figure 3. Determini
ng pea
ks from a sig
nal. (a) O
r
igin
al sign
al befo
r
e noi
se remo
val. (b) Signal
after noise re
moval usin
g an ada
ptive noise
-removal
filter. (c)
Hist
ogra
m
of respirato
r
y sign
a
l
after noise re
moval; baseli
ne is dete
r
mi
ned by most
occurrin
g value. Points tha
t
are highe
r than
baseline multi
p
lied a pa
ram
e
ter p are ma
rke
d
as
p
e
a
k
. (d) Valleys a
r
e dete
c
ted u
s
ing regio
nal
minima. The
detecte
d valleys are
circle
d
T
x
S
p
a
tio
te
m
p
o
r
a
l
o
b
ta
in
e
d
f
r
o
m
th
e
c
o
lu
m
n
x
o
f
s
u
bs
e
que
nt
f
r
a
m
e
s
fr
a
m
e
s
Y
fr
a
m
es
Y
fr
a
m
e
s
Y
50
10
fr
a
m
e
s
Y
50
10
50
10
Y
Fr
e
q
.
fr
a
m
es
Y
50
10
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 2, May 2015 : 329 – 334
332
In gene
ral, a re
spiratory
pattern con
s
ist
s
of one
peak
and
one valley. A semi-
automatically peak a
nd va
lley detection
was p
r
o
p
o
s
ed in [12]. Although thi
s
p
r
opo
se
d met
hod
wa
s able to
detect p
e
a
k
s and valleys
from a respir
atory pattern, the re
spirato
r
y sign
al is
n
o
t
obtaine
d fro
m
time-seq
ue
ntial imag
es.
It is di
re
ct
ly measured
by a di
gital volt
age
sig
nal
using
a
pre
s
sure se
n
s
or. Mo
reove
r
, manual u
s
e
r
revie
w
is al
so requi
re
d to verify the results.
In this study we propo
se a
n
automatic p
eak
an
d valley detection from respiratory signal
obtaine
d fro
m
diaph
ragm
motion (Figu
r
e 3). We first perform noi
se filtering u
s
ing an ad
apti
v
e
noise-remova
l
filter. Next step is to
set
a bas
eline v
a
lue ba
se
d o
n
the statisti
cal mode (mo
s
t
freque
ntly occurrin
g value) of the signal. A paramete
r
,
p
is used t
o
determin
e
the height of the
pea
k. Points
in the respi
r
a
t
ory sign
al th
at are
highe
r than the b
a
s
elin
e multipl
i
ed with
p
are
marked
as p
eak. T
he
si
milar p
r
o
c
e
s
s i
s
al
so
do
ne to d
e
tect
the valleys.
Instea
d of fi
nding
statistical mo
de, re
gional
minima of th
e sig
nal
a
r
e
cal
c
ulate
d
an
d multiplied
b
y
a paramete
r
,
v
.
All points bel
ow this valu
e are ma
rked a
s
valley.
3. Results
We teste
d
th
e pro
p
o
s
ed
method to a
total of 15 su
bject
s
(8 h
e
a
l
thy volunteers and
7
COPD
patien
t
s). The n
u
m
ber of fram
e
for each su
bject is 1
200
frames. Ta
b
l
e 1 sho
w
s the
numbe
r of re
spiratory patt
e
rn
s found a
n
d
the
numbe
r frame for on
e respiratory cycle.
Table 1. Nu
m
ber of re
spi
r
a
t
ory pattern
s found
a
nd the
averag
e num
ber of fram
e required for
one re
spi
r
ato
r
y cycle in he
althy volunteers
Subject #
Resp.
Pat.
#Frame/c
y
c
le
Health
y
Voluntee
rs
1 32
36.7
2 36
33.3
3 37
32.4
4 52
23.1
5 19
63.2
6 57
21.0
7 48
25.0
8 34
35.3
COPD p
a
tients
1 49
24.5
2 77
15.6
3 34
35.3
4 38
31.6
5 61
19.7
6 56
21.4
7 46
26.1
Ideally, the n
u
mbe
r
of fra
m
e for
one
respi
r
at
o
r
y cy
cle
ran
g
e
s
from 25
-35 f
r
a
m
e/cycle.
Figure 4
sh
o
w
s an
exam
ple of
re
spira
t
ory pattern
s
whi
c
h
su
ccessfully extra
c
ted fro
m
he
althy
volunteer #
3
. The numb
e
r of detected
respirator
y p
a
tterns i
s
37
and the nu
m
ber of fram
e per
cycle i
s
32.4
whi
c
h is
con
s
idere
d
as n
o
rmal re
spiratory motion.
Figure 4. Detected respiratory pattern
s for healthy vol
unteer #
3
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Autom
a
tic Extraction of
Diaphragm
Motion and
Re
spi
r
atory Pattern
from
… (Win
d
ra Swasti
ka)
333
Ho
wever, b
r
eathing irre
g
u
larity is a factor
th
at make
s re
spi
r
at
ory pattern e
x
traction
failed. Another facto
r
tha
t
affects the numbe
r
of detecte
d re
spirato
r
y pattern is respirat
ory
freque
ncy. F
o
r exam
ple,
subj
ect 6
ha
s the l
a
rg
est
numb
e
r of
extracted
re
spirato
r
y patte
rns
among th
e other h
ealthy voluntee
rs. Th
e subj
ect’
s
2
D
sp
atio tem
poral
sh
ows that this subje
c
t
has hi
gh re
sp
iratory freq
ue
ncy (Fig
ure 5
(
a)).
The h
ealthy
volunteer wh
o ha
s the
sm
allest
n
u
mbe
r
of dete
c
ted
respi
r
ato
r
y pat
terns is
subj
ect 5. A
s
we
can
se
e
in the Fig
u
re 5(b), subje
c
t 5 ha
s
sev
e
ral i
rre
gula
r
bre
a
thing
cy
cle
s
(pointe
d
by white arrows) t
hat make
the
s
y
s
t
em failed
to extrac
t them.
(a)
(b)
Figure 5. An example of (a
) high respirat
ory
frequ
en
cy and (b
) irreg
u
lar b
r
eathi
ng
of healthy
volunteers
For
COP
D
p
a
tients, the
n
u
mbe
r
of
extracte
d
re
spiratory patte
rn
s ten
d
s to b
e
hig
her
comp
ared wit
h
he
althy
vol
unteers. Figu
re 6 sh
o
w
s t
w
o
exampl
es of
COPD p
a
t
ients 5
a
nd
6.
The fra
m
e/cy
cle of the
s
e
p
a
tients a
r
e 1
9
.
7 and 2
1
.4, resp
ectively. It indicates th
a
t
these p
a
tien
ts
have smalle
r
lung volum
e
cap
a
city com
pare
d
with
he
althy subj
ect
s
. Several irre
gular breathin
g
s
were also fo
und in the first 100 fram
e
s
of the
pati
ent 6 (Figu
r
e
6(b)) an
d they failed to be
extracted.
(a)
(b)
Figure 6. Two example
s
o
f
extracted re
sp
iratory patt
e
rn
s of COP
D
patient 5 a
nd 6
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 2, May 2015 : 329 – 334
334
4. Conclusio
n
This study
propo
sed
an
a
u
tomatic
met
hod to
extra
c
t diaph
rag
m
motion a
nd
respi
r
ato
r
y
pattern
s from
time seq
uen
tial MR imag
es in
sagittal
plane. Ou
r
method
su
ccessfully extra
c
ts
diaph
rag
m
motion and re
spiratory patt
e
rn
s fo
r both
healthy volunteers and
COPD p
a
tien
ts.
Ho
wever, o
u
r study ha
s
certain limitati
ons. Fi
rs
t, it fails to dete
c
t irreg
u
lar
breathing
patte
rns
whi
c
h
ca
n o
c
cur du
rin
g
M
R
I a
c
qui
sition
. Seco
nd,
the
re
sult
s of th
e
prese
n
t
stud
y we
re
obtain
ed
from a small
numbe
r of subje
c
ts. La
rg
er nu
mbe
r
of
subje
c
t
s
for both healthy
volunteer
a
n
d
COPD
patien
t
s are requi
re
d to validate our meth
od.
Ackn
o
w
l
e
dg
ements
This
study wa
s su
ppo
rted i
n
part by MEXT Kakenhi
No
s. 2210
35
04 and 2
410
3
703.
Referen
ces
[1]
Shib
ata H, I
w
a
s
a
w
a T
,
Gotoh
T
,
Kagei S, Shin
ohar
a T
,
Og
ura T
,
Inoue T
.
Automatic T
r
a
cking of th
e
Respir
ator
y
M
o
tion of Lun
g Pa
rench
y
m
a
on D
y
n
a
mic
Mag
net
ic Reson
anc
e Imagi
ng: Comp
ariso
n
W
i
th
Pulmo
nar
y F
u
nction T
e
sts in Patie
n
ts W
i
th Chro
nic Ob
structive Pulm
onar
y D
i
seas
e
.
Journal of
thoracic i
m
agi
n
g
. 2012; 2
7
(6): 387-
392.
[2]
Diame
nt MJ,
Boech
a
t MI, Kang
arlo
o H. Real
‐
time sector
ultrasou
nd in
the eval
uatio
n of suspect
e
d
abn
ormal
i
ties o
f
diaphr
agm
atic motion.
Jour
nal of cli
n
ica
l
ul
trasoun
d
. 198
5
;
13(8): 539-5
4
3
.
[3]
Gerscovich E
O, Cronan M, McGahan JP, Jain
K, Jones CD, McDona
ld C. Ultrason
ogra
p
h
i
c
eval
uatio
n of di
aphr
agmatic m
o
tion.
Jour
na
l of ultrasou
nd i
n
me
dic
i
ne
. 2
0
01; 20(6): 5
97-
604.
[4]
Boussuges A,
Gole Y, Blanc
P. Diaphragm
atic motion studied by
m-mode ultrasonogr
aphy
.
C
hest
.
200
9; 135: 39
1
-
400.
[5]
Kotani T
,
Mina
mi S, T
a
kahashi K, Isobe K,
Nakata
Y, T
a
kaso M, Mori
ya
H. An an
al
ysis
of chest
w
a
l
l
and d
i
a
phra
g
m
motions in
patie
nts
w
i
t
h
i
d
io
pathic sc
oli
o
sis usi
ng d
y
n
a
mic bre
a
thin
g
MRI.
Spine
.
200
4; 29(3): 29
8-30
2.
[6]
Kolar
P, Ne
u
w
irth J, Sa
nd
a
J,
Sucha
nek
V, Svata Z
,
V
o
lej
n
ik
J,
Piv
e
c M. Ana
l
ysis
of di
ap
hrag
m
moveme
nt dur
i
ng ti
dal
bre
a
thi
ng
and
d
u
rin
g
i
t
s activatio
n
w
h
ile
br
eath
ho
l
d
in
g us
ing
MRI
s
y
nchr
oniz
e
d
w
i
t
h
spir
ometr
y
.
Physiological Research
. 20
0
9
; 58(3): 38
3.
[7]
I
w
as
a
w
a
T
,
T
a
kah
a
shi
H,
Ogura T
,
Asakura A,
Got
o
h
T
,
Shibata
H
,
Inoue
T
.
Influenc
e
of th
e
distrib
u
tion of emph
ys
ema o
n
dia
phra
g
mati
c moti
on in pa
tients
w
i
t
h
chr
onic o
b
structiv
e pulmo
nar
y
dise
ase.
Jap
a
n
e
se jo
urna
l of radi
olo
g
y
. 201
1; 29(4): 256-
26
4.
[8]
Otsu N. A threshold selection me
th
od
fro
m
gray
-l
e
v
el
hi
sto
g
ra
ms.
Autom
a
tica
. 19
75; 1
1
(2
85-2
96): 2
3
-
27.
[9]
Cha
n
T
F
, Vese LA. Active contours
w
i
t
h
o
u
t
edges. Imag
e Processi
ng,
IEEE Transactions
. 20
01;
10(2): 26
6-2
7
7
.
[10]
Alfians
ya
h A.
A Unifi
e
d
En
e
r
g
y
Ap
proac
h
fo
r B-Sp
lin
e
Snake
in
Me
d
i
cal Ima
g
e
Se
gmentati
on.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2010; 8(2): 1
7
5
-18
6
.
[11]
T
j
andrasa H, W
ija
yant
i
A, S
u
ciati
N. Se
g
m
entat
io
n
of the r
e
tina
l
opti
c
nerv
e
h
e
a
d
usi
ng
Ho
ug
h
transform and
active conto
u
r
models.
T
E
L
K
OMNIKA Indones
ian J
ourn
a
l of Electric
al
Engin
eer
ing
.
201
2; 10(3): 53
1-53
6.
[12]
Lu W
,
N
y
stro
m MM, Parikh
PJ, F
ooshe
e
DR,
Hu
be
nsc
h
midt JP, Bra
d
le
y J
D
, Lo
w
DA. A semi-
automatic m
e
thod for
peak
a
nd
va
lle
y
dete
c
tion i
n
free-br
eathi
ng res
p
ir
a
t
or
y
w
a
veform
s.
Med. Phys.
200
6; 333
63
4(6).
Evaluation Warning : The document was created with Spire.PDF for Python.