I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
1
6
,
No
.
3
,
J
u
n
e
20
2
6
,
p
p
.
1588
~
1
6
0
7
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
1
6
i
3
.
pp
1
5
8
8
-
1
6
0
7
1588
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
H
y
brid
d
eep l
ea
rning
(ILeS
-
Net
)
f
o
r
lung
ca
ncer
cla
ss
ificatio
n
in cloud
-
Io
T
hea
lt
hca
re sy
stems
Af
f
ro
s
e
1
,
Cheruk
u Sa
nd
esh
K
um
a
r
1
,
Arc
hek
P
r
a
v
ee
n K
um
a
r
2
1
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
A
m
i
t
y
U
n
i
v
e
r
si
t
y
R
a
j
a
s
t
h
a
n
,
Ja
i
p
u
r
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
G
a
t
e
s
I
n
st
i
t
u
t
e
o
f
T
e
c
h
n
o
l
o
g
y
,
A
n
a
n
t
a
p
u
r
a
m
u
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
8
,
2
0
2
5
R
ev
is
ed
Ap
r
2
,
2
0
2
6
Acc
ep
ted
Ap
r
2
6
,
2
0
2
6
T
h
is
s
t
u
d
y
p
re
se
n
ts
a
c
l
o
u
d
–
I
n
ter
n
e
t
o
f
T
h
in
g
s
(c
l
o
u
d
-
I
o
T)
b
a
se
d
in
te
ll
i
g
e
n
t
d
e
c
is
io
n
s
u
p
p
o
rt
fr
a
m
e
w
o
r
k
f
o
r
l
u
n
g
c
a
n
c
e
r
c
la
ss
i
fi
c
a
t
i
o
n
a
n
d
trea
tme
n
t
re
c
o
m
m
e
n
d
a
t
i
o
n
,
c
e
n
te
re
d
o
n
a
h
y
b
r
i
d
d
e
e
p
lea
r
n
i
n
g
m
o
d
e
l
te
rm
e
d
I
Le
S
-
N
e
t
.
Co
m
p
u
ted
to
m
o
g
ra
p
h
y
(CT)
ima
g
e
s
fro
m
a
b
e
n
c
h
m
a
rk
d
a
tas
e
t
a
re
first
p
re
p
ro
c
e
ss
e
d
u
si
n
g
G
a
u
ss
ian
fil
t
e
rin
g
t
o
e
n
h
a
n
c
e
ima
g
e
q
u
a
li
t
y
.
Ca
n
c
e
ro
u
s
re
g
io
n
s
a
re
id
e
n
ti
f
ied
u
si
n
g
a
n
I
m
p
ro
v
e
d
BIRCH
(I
-
BIRCH)
se
g
m
e
n
tatio
n
a
p
p
ro
a
c
h
,
f
o
ll
o
we
d
b
y
fe
a
tu
re
e
x
trac
ti
o
n
u
si
n
g
sh
a
p
e
d
e
sc
rip
t
o
rs,
c
o
lo
r
fe
a
tu
re
s,
a
n
d
Im
p
ro
v
e
d
lo
c
a
l
G
a
b
o
r
XO
R
p
a
tt
e
r
n
(I
-
LG
XP)
tex
t
u
re
s.
T
h
e
e
x
trac
ted
fe
a
tu
re
s
a
re
c
las
sified
u
sin
g
ILeS
-
Ne
t,
wh
ic
h
in
teg
ra
tes
Im
p
ro
v
e
d
LeN
e
t
-
5
a
n
d
S
q
u
e
e
z
e
Ne
t
a
r
c
h
it
e
c
tu
re
s
to
a
c
h
iev
e
imp
ro
v
e
d
c
la
ss
ifi
c
a
ti
o
n
p
e
rfo
rm
a
n
c
e
with
re
d
u
c
e
d
o
v
e
rfi
t
ti
n
g
.
Ba
se
d
o
n
th
e
c
las
sifica
ti
o
n
r
e
su
lt
s,
th
e
fra
m
e
wo
rk
p
ro
v
i
d
e
s
su
p
p
o
rti
v
e
re
c
o
m
m
e
n
d
a
ti
o
n
s
to
a
ss
ist
c
li
n
ica
l
d
e
c
isio
n
-
m
a
k
in
g
.
E
x
p
e
rime
n
tal
re
su
lt
s d
e
m
o
n
stra
te
th
a
t
th
e
p
r
o
p
o
se
d
ILeS
-
Ne
t
m
o
d
e
l
a
c
h
iev
e
s
a
m
a
x
imu
m
a
c
c
u
ra
c
y
o
f
0
.
9
5
1
,
o
u
tp
e
rf
o
rm
in
g
se
v
e
ra
l
c
o
n
v
e
n
t
io
n
a
l
a
n
d
sta
te
-
of
-
t
h
e
-
a
rt
m
e
th
o
d
s.
T
h
e
c
lo
u
d
–
I
o
T
i
n
teg
ra
ti
o
n
f
u
rt
h
e
r
e
n
a
b
les
sc
a
lab
le,
re
a
l
-
ti
m
e
,
a
n
d
se
c
u
re
d
a
t
a
p
ro
c
e
ss
in
g
,
h
ig
h
li
g
h
ti
n
g
t
h
e
fra
m
e
wo
rk
’s
p
o
ten
ti
a
l
fo
r
p
ra
c
ti
c
a
l
c
o
m
p
u
ter
-
a
id
e
d
l
u
n
g
c
a
n
c
e
r
d
iag
n
o
sis
.
K
ey
w
o
r
d
s
:
L
ung
ca
n
ce
r
class
if
icatio
n
I
-
B
I
R
C
H
I
-
L
GXP
I
m
p
r
o
v
ed
I
L
eNe
t
-
5
(
I
L
eNe
t
-
5)
C
lo
u
d
-
I
o
T
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
:
Af
f
r
o
s
e
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
,
Am
ity
Un
iv
er
s
ity
R
ajasth
an
J
aip
u
r
,
R
ajasth
an
3
0
3
0
0
2
,
I
n
d
i
a
E
m
ail:
r
o
s
ea
f
f
r
o
s
e5
5
7
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
R
ec
en
t
tech
n
o
lo
g
ical
ad
v
an
c
em
en
ts
h
av
e
b
r
o
u
g
h
t
ab
o
u
t
im
p
o
r
tan
t
ch
an
g
es
ac
r
o
s
s
n
u
m
er
o
u
s
in
d
u
s
tr
ies,
in
clu
d
in
g
h
ea
lth
ca
r
e
[
1
]
.
Par
ticu
lar
ly
im
p
o
r
tan
t is th
e
in
co
r
p
o
r
atio
n
o
f
clo
u
d
c
o
m
p
u
tin
g
[
2
]
a
n
d
th
e
I
n
ter
n
et
o
f
T
h
i
n
g
s
(
I
o
T
)
in
t
o
h
ea
lth
ca
r
e
s
y
s
tem
s
[
3
]
–
[
5
]
,
wh
ich
h
as
led
to
th
e
d
ev
elo
p
m
en
t
o
f
ad
v
an
ce
d
d
ec
is
io
n
s
u
p
p
o
r
t
s
y
s
tem
s
.
T
h
i
s
m
er
g
in
g
o
f
tech
n
o
l
o
g
ies
s
h
o
ws
g
r
ea
t
p
o
ten
tial
in
tr
a
n
s
f
o
r
m
in
g
p
atien
t
ca
r
e,
im
p
r
o
v
in
g
d
iag
n
o
s
tic
p
r
ec
is
io
n
,
an
d
s
tr
ea
m
lin
in
g
h
ea
lth
ca
r
e
o
p
er
atio
n
s
.
At
th
e
s
am
e
tim
e
,
in
ter
n
et
o
f
t
h
in
g
s
(
I
o
T
)
d
ev
ices
s
u
ch
as
wea
r
ab
le
f
itn
ess
tr
ac
k
er
s
,
m
ed
ical
s
e
n
s
o
r
s
,
an
d
in
tellig
en
t
m
o
n
ito
r
in
g
d
e
v
ices
en
ab
le
r
em
o
te
p
atien
t
m
o
n
ito
r
in
g
,
r
e
al
-
tim
e
d
ata
g
ath
e
r
in
g
,
an
d
p
r
o
m
p
t
in
ter
v
en
tio
n
s
,
t
h
er
eb
y
p
r
o
m
o
tin
g
a
p
atie
n
t
-
ce
n
ter
ed
ap
p
r
o
ac
h
to
h
ea
lth
ca
r
e
d
eliv
er
y
[
6
]
–
[
8
]
.
Desp
ite
th
e
p
r
o
g
r
ess
in
h
ea
lth
ca
r
e
tech
n
o
lo
g
y
,
lu
n
g
ca
n
c
er
r
em
ain
s
a
s
ig
n
if
ican
t
g
lo
b
al
h
ea
lth
co
n
ce
r
n
,
with
alar
m
i
n
g
r
ates
o
f
in
cid
en
ce
an
d
m
o
r
tality
.
A
n
n
u
ally
,
m
o
r
e
th
an
1
.
8
m
illi
o
n
n
ew
ca
s
es
o
f
lu
n
g
ca
n
ce
r
ar
e
id
e
n
tifie
d
wo
r
ld
w
id
e,
lead
in
g
to
1
.
4
m
illi
o
n
d
ea
th
s
[
9
]
,
[
1
0
]
.
L
ung
ca
n
ce
r
h
o
ld
s
th
e
h
ig
h
est
m
o
r
tality
r
ate
am
o
n
g
v
ar
i
o
u
s
ty
p
es
o
f
ca
n
ce
r
,
h
ig
h
lig
h
tin
g
th
e
cr
itical
n
ec
ess
ity
f
o
r
ef
f
i
cien
t
d
etec
tio
n
a
n
d
tr
ea
tm
en
t a
p
p
r
o
ac
h
es
[
1
1
]
,
[
1
2
]
.
Ack
n
o
wled
g
in
g
th
e
d
i
f
f
icu
ltie
s
lin
k
ed
to
d
iag
n
o
s
in
g
lu
n
g
ca
n
ce
r
,
esp
ec
ially
th
e
lack
o
f
s
y
m
p
to
m
s
in
its
ea
r
ly
p
h
ases
,
alo
n
g
s
id
e
f
ac
to
r
s
lik
e
r
esp
ir
ato
r
y
p
r
o
b
le
m
s
an
d
s
m
o
k
in
g
h
is
to
r
y
,
h
i
g
h
lig
h
ts
th
e
cr
u
cial
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:
2088
-
8
7
0
8
Hyb
r
id
d
ee
p
lea
r
n
in
g
(
I
LeS
-
N
et)
fo
r
lu
n
g
ca
n
ce
r
cla
s
s
ifica
tio
n
in
…
(
A
ffr
o
s
e
)
1589
s
ig
n
if
ican
ce
o
f
ea
r
ly
d
etec
tio
n
.
C
o
n
v
en
tio
n
al
d
iag
n
o
s
tic
im
ag
in
g
m
eth
o
d
s
s
u
ch
as
co
m
p
u
ted
to
m
o
g
r
a
p
h
y
(
CT
)
,
p
o
s
itro
n
em
is
s
io
n
t
o
m
o
g
r
ap
h
y
(
PET
)
,
an
d
m
ag
n
etic
r
eso
n
an
ce
im
ag
i
n
g
(
MRI
)
f
ac
e
co
n
s
tr
ain
ts
in
id
en
tify
in
g
l
ung
m
ass
es,
p
r
o
m
p
tin
g
th
e
e
x
p
lo
r
atio
n
o
f
in
n
o
v
ativ
e
tech
n
o
lo
g
ies to
f
ill th
ese
g
ap
s
[
1
3
]
,
[
1
4
]
.
T
o
ta
ck
le
th
ese
c
h
al
le
n
g
es
,
s
cie
n
tis
ts
h
a
v
e
p
r
o
g
r
ess
i
v
el
y
e
m
b
r
a
ce
d
ar
ti
f
ic
ial
i
n
te
lli
g
e
n
c
e
(
A
I
)
a
n
d
m
ac
h
i
n
e
lea
r
n
i
n
g
(
M
L
)
m
e
th
o
d
o
l
o
g
ies
t
o
cr
ea
te
i
n
v
e
n
ti
v
e
s
o
lu
t
io
n
s
f
o
r
d
ete
cti
n
g
l
u
n
g
c
an
c
er
.
AI
’
s
e
x
p
e
r
tis
e
in
co
m
p
u
te
r
v
is
i
o
n
,
b
i
g
d
a
ta
a
n
a
l
y
s
is
,
an
d
h
ea
l
th
ca
r
e
u
til
iz
ati
o
n
p
o
s
i
ti
o
n
s
it
f
a
v
o
r
ab
ly
f
o
r
ta
c
k
li
n
g
class
if
ica
ti
o
n
is
s
u
es
i
n
m
e
d
i
ca
l i
m
a
g
i
n
g
[
1
5
]
.
W
it
h
i
n
t
h
is
f
r
a
m
ew
o
r
k
,
a
cl
o
u
d
-
I
o
T
d
ec
is
io
n
s
u
p
p
o
r
t s
y
s
te
m
b
as
ed
o
n
D
L
[
1
6
]
,
[
1
7
]
em
e
r
g
es
as
a
co
m
p
el
li
n
g
s
tr
a
te
g
y
t
o
e
n
h
a
n
c
e
lu
n
g
ca
n
c
er
d
ete
cti
o
n
a
n
d
a
d
v
a
n
ce
p
ati
en
t
o
u
tc
o
m
es
.
Utilizin
g
AI
,
clo
u
d
co
m
p
u
tin
g
,
an
d
I
o
T
tech
n
o
lo
g
ies,
th
is
s
y
s
tem
aim
s
to
ex
ce
ed
th
e
co
n
s
tr
ain
ts
o
f
co
n
v
en
tio
n
al
d
ia
g
n
o
s
tic
tech
n
iq
u
es,
eq
u
ip
p
in
g
h
ea
lth
ca
r
e
p
r
o
v
id
er
s
with
in
s
tan
t
in
s
i
g
h
ts
an
d
p
r
ac
tical
r
ec
o
m
m
en
d
atio
n
s
f
o
r
ea
r
ly
d
e
tectio
n
an
d
in
ter
v
e
n
tio
n
.
T
h
is
r
esear
ch
o
u
tlin
es
th
e
s
tr
u
ctu
r
e
o
f
a
d
ee
p
lear
n
in
g
-
b
ased
clo
u
d
-
I
o
T
d
ec
is
io
n
s
u
p
p
o
r
t
s
y
s
tem
d
esig
n
ed
f
o
r
lu
n
g
ca
n
ce
r
class
if
icatio
n
.
As
a
r
esu
lt,
th
e
p
r
o
p
o
s
ed
m
o
d
el
p
r
esen
ts
f
o
u
r
u
n
iq
u
e
co
n
tr
ib
u
tio
n
s
,
ea
c
h
m
eth
o
d
ically
ex
p
lain
ed
as f
o
llo
ws.
−
Pro
p
o
s
in
g
I
-
B
I
R
C
H
to
p
r
o
v
id
e
a
n
ef
f
ec
tiv
e
s
eg
m
e
n
tat
io
n
v
ia
in
c
o
r
p
o
r
atin
g
au
to
m
ated
th
r
esh
o
ld
in
itializatio
n
an
d
im
p
r
o
v
e
d
d
is
tan
ce
m
atr
ices
f
o
r
e
n
h
an
ci
n
g
th
e
clu
s
ter
in
g
ac
cu
r
ac
y
,
m
e
m
o
r
y
u
tili
za
tio
n
,
an
d
tim
e
co
m
p
lex
ity
b
y
ef
f
ec
ti
v
ely
ad
ap
tin
g
to
e
x
ten
s
iv
e
d
at
asets
.
−
E
x
tr
ac
tin
g
I
-
L
GXP
-
b
ased
f
e
atu
r
es
in
ad
d
itio
n
to
c
o
n
v
e
n
tio
n
al
s
h
ap
e
an
d
c
o
lo
r
f
ea
tu
r
es
f
r
o
m
th
e
s
eg
m
en
ted
im
ag
e
to
p
r
o
v
i
d
e
a
n
ac
cu
r
ate
l
u
n
g
ca
n
ce
r
class
if
icatio
n
.
T
h
is
I
-
L
GXP
is
an
e
n
h
an
ce
d
f
ea
tu
r
e,
wh
er
e,
th
e
p
a
r
am
eter
s
ar
e
ad
ju
s
ted
wh
ich
en
h
a
n
ce
s
th
e
f
i
lter
’
s
ab
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
v
ar
io
u
s
tex
tu
r
e
p
atter
n
s
in
im
ag
es,
im
p
r
o
v
in
g
o
v
er
all
im
ag
e
q
u
ality
an
d
an
aly
s
is
ac
cu
r
ac
y
.
−
Pro
p
o
s
in
g
I
L
eS
-
Net
m
o
d
el,
wh
ich
is
th
e
h
y
b
r
id
izatio
n
o
f
I
L
eNe
t
-
5
an
d
Sq
u
ee
ze
Net
m
o
d
els
to
p
r
o
v
id
e
b
etter
lu
n
g
ca
n
ce
r
class
if
icatio
n
.
Her
e
,
I
L
eNe
t
-
5
m
o
d
el
is
th
e
co
n
v
en
tio
n
al
L
eNe
t
-
5
m
o
d
el’
s
im
p
r
o
v
ed
v
er
s
io
n
,
wh
er
e
th
e
m
o
d
el’
s
s
tab
ilit
y
an
d
ef
f
icien
cy
ar
e
en
h
an
ce
d
d
u
r
i
n
g
tr
ain
i
n
g
v
i
a
in
co
r
p
o
r
atin
g
I
m
p
r
o
v
ed
B
N
lay
er
a
n
d
Har
d
E
lis
h
Swis
h
m
ax
o
u
t a
ctiv
atio
n
f
u
n
ctio
n
.
T
h
e
r
est
o
f
th
is
s
tu
d
y
is
co
n
s
tr
u
cted
as
f
o
llo
ws:
Sectio
n
2
r
ev
iews
o
n
p
r
ev
io
u
s
s
tu
d
y
with
in
th
e
d
o
m
ain
o
f
lu
n
g
ca
n
ce
r
.
Fo
llo
win
g
th
at,
s
ec
tio
n
3
ex
p
l
o
r
es
t
h
e
p
r
o
ce
s
s
o
f
t
h
e
p
r
o
p
o
s
ed
I
L
eS
-
Net
f
r
am
ewo
r
k
.
Sectio
n
4
p
r
esen
ts
th
e
ev
al
u
atio
n
f
in
d
in
g
s
an
d
p
r
esen
ts
a
co
m
p
r
eh
en
s
iv
e
d
is
cu
s
s
io
n
.
Fin
ally
,
s
ec
tio
n
5
en
ca
p
s
u
lates th
e
s
tu
d
y
’
s
f
in
d
in
g
s
an
d
c
o
n
clu
s
io
n
s
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
T
h
is
s
e
cti
o
n
p
r
o
v
i
d
es
a
c
o
n
cis
e
o
v
er
v
i
ew
o
f
l
u
n
g
c
an
ce
r
cla
s
s
if
i
ca
t
io
n
b
y
s
y
n
t
h
esi
zi
n
g
i
n
s
ig
h
ts
f
r
o
m
ten
r
el
ev
a
n
t
r
es
ea
r
c
h
p
a
p
er
s
.
I
n
2
0
2
3
,
F
ar
u
q
u
i
et
a
l
.
[
1
8
]
h
as
in
t
r
o
d
u
ce
d
h
e
alt
h
c
a
r
e
-
as
-
a
-
s
e
r
v
ic
e
(
HA
AS)
m
o
d
el
in
s
p
i
r
e
d
b
y
s
o
f
tw
ar
e
-
as
-
a
-
s
e
r
v
i
ce
(
SA
AS)
wit
h
i
n
t
h
e
cl
o
u
d
c
o
m
p
u
t
in
g
p
ar
a
d
i
g
m
.
I
n
2
0
2
3
,
T
o
m
ass
i
n
i
et
a
l.
[
1
9
]
h
a
v
e
d
e
v
el
o
p
e
d
a
L
UC
Y
a
d
v
a
n
c
ed
o
n
-
c
lo
u
d
d
ec
is
i
o
n
s
u
p
p
o
r
t
s
y
s
t
em
f
r
o
m
t
h
o
r
a
x
C
T
s
c
a
n
s
.
I
n
2
0
2
0
,
Hw
an
g
et
a
l
.
[
2
0
]
h
as
c
o
m
p
a
r
e
d
t
h
e
C
T
i
n
t
er
p
r
et
ati
o
n
b
ef
o
r
e
a
n
d
a
f
te
r
i
m
p
le
m
e
n
ta
ti
o
n
o
f
a
c
o
m
p
u
te
r
i
ze
d
s
y
s
t
em
f
o
r
lu
n
g
n
o
d
u
l
e
d
ete
cti
o
n
.
I
n
2
0
1
9
,
M
aso
o
d
et
a
l
.
[
2
1
]
h
a
v
e
d
ev
el
o
p
e
d
a
3
D
DC
NN
f
o
r
l
u
n
g
n
o
d
u
le
d
et
ec
t
io
n
b
as
ed
o
n
ass
is
ti
n
g
th
e
r
a
d
i
o
l
o
g
is
ts
.
I
n
2
0
2
2
,
Kas
in
at
h
a
n
a
n
d
J
a
y
ak
u
m
a
r
[
2
2
]
h
as
p
r
es
en
te
d
a
C
l
o
u
d
-
L
T
DSC
a
h
y
b
r
i
d
t
ec
h
n
i
q
u
e
f
o
r
PE
T
/C
T
i
m
a
g
es
cl
ass
if
y
i
n
g
a
n
d
v
al
id
ati
n
g
d
i
f
f
e
r
e
n
t
s
ta
g
es
o
f
l
u
n
g
t
u
m
o
r
p
r
o
g
r
ess
i
o
n
.
A
n
o
v
el
SC
M
O
-
ML
L
2
C
m
e
th
o
d
was
in
tr
o
d
u
ce
d
b
y
Va
ll
u
r
u
a
n
d
J
e
y
a
[
2
3
]
i
n
2020
f
o
r
u
s
e
w
it
h
C
T
s
c
a
n
s
.
I
n
2
0
2
0
,
Ma
s
o
o
d
e
t
a
l.
[
2
4
]
h
as
s
u
g
g
este
d
an
e
n
h
a
n
ce
d
m
R
F
C
N
b
as
e
d
a
u
t
o
m
at
ed
d
ec
is
io
n
s
u
p
p
o
r
t
s
y
s
t
em
f
o
r
lu
n
g
n
o
d
u
le
d
et
ec
t
io
n
a
n
d
cl
ass
if
ica
ti
o
n
.
Ut
ili
zi
n
g
m
R
FC
N
as
th
e
im
a
g
e
cl
ass
i
f
ie
r
b
a
ck
b
o
n
e
f
o
r
f
ea
t
u
r
e
ex
t
r
ac
ti
o
n
a
n
d
m
L
R
PN
wit
h
PS
SM
was
in
te
g
r
al
.
A
d
d
iti
o
n
all
y
,
a
d
ec
o
n
v
o
lu
ti
o
n
al
l
a
y
e
r
was
i
n
t
r
o
d
u
c
ed
t
o
in
c
o
r
p
o
r
a
te
t
h
e
s
u
g
g
est
ed
m
L
R
PN
i
n
t
o
t
h
is
f
r
am
ew
o
r
k
,
f
ac
il
itati
n
g
t
h
e
au
to
m
a
tic
s
ele
cti
o
n
o
f
p
o
t
en
tial
R
O
I
.
I
n
2
0
2
1
,
Mis
h
r
a
et
a
l
.
[
2
5
]
h
a
v
e
d
e
v
e
lo
p
ed
a
s
u
s
t
ai
n
a
b
l
e
lu
n
g
ca
n
ce
r
d
et
ec
t
io
n
m
o
d
e
l
t
o
in
te
g
r
at
e
th
e
I
o
H
T
a
n
d
co
m
p
u
tati
o
n
a
l i
n
t
ell
ig
en
ce
,
ca
u
s
i
n
g
t
h
e
le
ast
h
ar
m
t
o
t
h
e
e
n
v
ir
o
n
m
e
n
t
.
I
n
2
0
2
3
,
R
a
za
et
a
l.
[
2
6
]
h
as
p
r
ese
n
te
d
a
tr
a
n
s
f
e
r
l
ea
r
n
in
g
(
T
L
)
b
ase
d
p
r
e
d
i
ct
o
r
ca
l
le
d
L
ung
-
E
f
f
N
et
f
o
r
l
u
n
g
ca
n
c
er
cl
ass
if
ic
ati
o
n
.
C
o
n
s
tr
u
cte
d
b
as
ed
o
n
th
e
E
f
f
ici
en
tNe
t
a
r
c
h
it
ec
t
u
r
e
,
it
w
as
f
u
r
t
h
e
r
e
n
h
a
n
c
ed
b
y
in
c
o
r
p
o
r
ati
n
g
a
d
d
it
io
n
al
to
p
la
y
er
s
i
n
t
h
e
class
i
f
i
ca
t
io
n
h
e
ad
o
f
th
e
m
o
d
el
.
I
n
2
0
2
4
,
Go
m
i
asti
e
t
a
l.
[
2
7
]
h
a
v
e
e
n
h
a
n
c
e
d
th
e
e
f
f
ici
en
cy
o
f
lu
n
g
c
an
ce
r
class
i
f
i
ca
t
io
n
p
e
r
f
o
r
m
a
n
c
e
v
ia
s
u
p
p
o
r
t
v
ec
to
r
m
a
ch
in
e
(
S
V
M)
wi
th
h
y
p
e
r
p
a
r
a
m
et
er
tu
n
i
n
g
.
I
n
2
0
2
4
,
A
m
i
n
et
a
l
.
[
2
8
]
h
as
p
r
o
p
o
s
e
d
m
u
lt
im
o
d
al
non
-
s
m
al
l
c
ell
l
u
n
g
c
an
ce
r
(
NSC
L
C
)
c
lass
i
f
i
ca
t
io
n
u
s
i
n
g
C
NN
m
o
d
el
.
Fro
m
t
h
e
g
e
n
o
m
i
c
d
at
a
c
o
m
m
o
n
s
(
GDC
)
p
o
r
tal
d
ata
was
g
at
h
e
r
e
d
an
d
t
h
en
d
iv
er
s
e
p
r
e
-
p
r
o
ce
s
s
i
n
g
te
c
h
n
iq
u
es
wer
e
u
tili
ze
d
f
o
r
p
r
e
p
a
r
i
n
g
t
h
e
g
e
n
o
m
ic
as
w
ell
as
s
l
id
e
i
m
a
g
es
.
I
n
2
0
2
4
,
N
o
m
a
n
et
a
l
.
[
2
9
]
h
as
p
r
o
p
o
s
e
d
L
ung
CT
-
NE
T
,
w
h
i
ch
was
a
T
L
-
b
ase
d
ar
c
h
it
ec
t
u
r
e
c
o
u
p
l
e
d
w
it
h
E
L
f
o
r
t
h
e
l
u
n
g
n
o
d
u
l
e
c
ate
g
o
r
iz
ati
o
n
.
I
n
2
0
2
5
,
Z
h
a
o
et
a
l
.
[
3
0
]
h
as
p
r
o
p
o
s
ed
Po
r
t
Net
,
w
h
i
ch
was
d
ev
el
o
p
e
d
f
o
r
r
e
d
u
ci
n
g
t
h
e
m
o
d
el
’
s
p
a
r
am
ete
r
s
i
ze
alo
n
g
wit
h
at
tai
n
i
n
g
t
h
e
li
g
h
t
wei
g
h
t
ch
a
r
ac
te
r
is
t
ics
wit
h
o
u
t
c
o
m
p
r
o
m
is
in
g
t
h
e
c
ate
g
o
r
iz
ati
o
n
ac
c
u
r
a
cy
.
I
n
2
0
2
3
,
S
u
n
et
a
l
.
[
3
1
]
p
r
o
p
o
s
e
d
Swi
n
T
r
a
n
s
f
o
r
m
er
f
o
r
cl
ass
i
f
i
ca
ti
o
n
as
well
as
s
e
g
m
e
n
t
ati
o
n
o
f
lu
n
g
ca
n
c
e
r
.
I
n
2
0
2
4
,
I
m
r
a
n
e
t
a
l.
[
3
2
]
p
r
o
p
o
s
e
d
C
NN
-
Vi
T
f
o
r
NSC
L
C
d
e
tec
ti
o
n
a
n
d
cl
ass
i
f
ic
ati
o
n
.
T
h
e
NSC
L
C
was
ca
t
eg
o
r
iz
e
d
i
n
t
o
n
o
r
m
al
,
s
q
u
a
m
o
u
s
c
ell
ca
r
ci
n
o
m
a
a
n
d
a
d
e
n
o
c
ar
ci
n
o
m
a
b
y
t
h
is
m
o
d
e
l.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
3
,
J
u
n
e
20
2
6
:
1
5
8
8
-
1
6
0
7
1590
3.
P
RO
P
O
SE
D
I
L
E
S
-
NE
T
P
O
WE
R
E
D
CL
O
UD
-
I
O
T
F
RA
M
E
WO
RK
F
O
R
I
NT
E
L
L
I
G
E
NT
L
UNG
CANC
E
R
DE
T
E
C
T
I
O
N
AN
D
RE
CO
M
M
E
NDA
T
I
O
N
S
Ar
is
in
g
f
r
o
m
t
h
e
u
n
c
o
n
tr
o
lled
g
r
o
wth
o
f
ab
n
o
r
m
al
ce
lls
in
lu
n
g
tis
s
u
es,
lu
n
g
ca
n
ce
r
s
tan
d
s
as
o
n
e
o
f
th
e
m
o
s
t
p
r
e
v
alen
t
an
d
d
ea
d
l
y
ty
p
es
o
f
ca
n
ce
r
wo
r
l
d
wid
e.
L
ung
ca
n
ce
r
class
if
icatio
n
aim
s
to
d
if
f
e
r
en
tiate
th
ese
s
u
b
ty
p
es
an
d
id
en
tify
s
p
ec
if
ic
ch
ar
ac
ter
is
tics
o
f
th
e
tu
m
o
r
,
s
u
ch
as
s
ize,
lo
ca
tio
n
,
an
d
h
is
to
lo
g
ical
f
ea
tu
r
es,
wh
ich
ca
n
i
n
f
o
r
m
tr
ea
tm
en
t
d
ec
is
io
n
s
an
d
p
r
o
g
n
o
s
is
.
C
o
n
v
en
tio
n
al
m
eth
o
d
s
f
o
r
lu
n
g
ca
n
ce
r
class
if
icatio
n
ty
p
ically
r
ely
o
n
h
is
to
p
ath
o
lo
g
ical
e
x
am
in
ati
o
n
o
f
tis
s
u
e
s
am
p
les
o
b
tain
e
d
th
r
o
u
g
h
in
v
asiv
e
p
r
o
ce
d
u
r
es
lik
e
b
io
p
s
ies
o
r
s
u
r
g
ical
r
esectio
n
.
B
u
t
th
ese
m
eth
o
d
s
m
ay
n
o
t
f
u
lly
ca
p
tu
r
e
th
e
m
o
lecu
lar
an
d
g
en
etic
co
m
p
lex
ities
o
f
lu
n
g
ca
n
ce
r
,
wh
ich
ar
e
i
n
cr
ea
s
in
g
ly
r
ec
o
g
n
ized
as
im
p
o
r
ta
n
t
f
ac
to
r
s
in
ca
n
ce
r
b
eh
av
io
r
a
n
d
tr
ea
tm
en
t
r
esp
o
n
s
e.
Ad
d
itio
n
ally
,
m
ed
ical
im
ag
in
g
tech
n
iq
u
es
s
u
ch
as
C
T
,
X
-
r
ay
s
,
an
d
MRI
ar
e
u
s
ed
to
v
is
u
alize
th
e
lu
n
g
an
d
d
etec
t
a
b
n
o
r
m
alities
.
R
ad
io
lo
g
is
ts
in
ter
p
r
et
th
ese
im
ag
es
to
id
en
tify
s
u
s
p
icio
u
s
lesi
o
n
s
,
ass
e
s
s
tu
m
o
r
s
ize
an
d
lo
ca
tio
n
,
an
d
g
u
id
e
tr
ea
tm
en
t p
lan
n
in
g
.
Acc
o
r
d
in
g
ly
,
I
L
eS
-
Net
p
o
wer
ed
clo
u
d
-
I
o
T
f
r
am
ewo
r
k
f
o
r
i
n
tellig
en
t
lu
n
g
ca
n
ce
r
d
etec
tio
n
a
n
d
r
ec
o
m
m
en
d
atio
n
s
is
p
r
o
p
o
s
ed
in
th
is
wo
r
k
.
I
n
itially
,
th
e
ac
q
u
i
r
ed
C
T
im
ag
es
ar
e
p
r
o
ce
s
s
ed
b
y
Gau
s
s
ian
f
ilter
in
g
in
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e.
Fo
llo
win
g
th
is
,
s
eg
m
en
tatio
n
u
s
in
g
I
-
B
I
R
C
H
is
em
p
lo
y
ed
t
o
s
eg
m
en
t
th
e
ca
n
ce
r
o
u
s
p
o
r
tio
n
f
r
o
m
th
e
p
r
e
-
p
r
o
ce
s
s
ed
C
T
im
ag
es.
T
h
en
,
th
e
s
eg
m
en
te
d
im
ag
es
a
r
e
s
u
b
jecte
d
to
f
ea
tu
r
e
ex
tr
a
ctio
n
s
tag
e
to
ac
q
u
ir
e
t
h
e
p
e
r
tin
en
t
f
ea
tu
r
es.
T
h
is
p
r
o
ce
d
u
r
e
in
cl
u
d
es
r
etr
iev
in
g
s
h
ap
e
f
ea
tu
r
es,
I
-
L
GXP,
an
d
co
lo
r
f
ea
tu
r
es
f
r
o
m
th
e
s
eg
m
en
ted
im
ag
e.
Nex
t,
th
e
r
etr
iev
ed
f
ea
tu
r
es
ar
e
f
ed
as
in
p
u
t
t
o
th
e
h
y
b
r
i
d
m
o
d
el,
c
o
m
p
r
is
in
g
I
L
eNe
t
-
5
an
d
Sq
u
ee
ze
Netis
p
r
esen
ted
f
o
r
l
u
n
g
ca
n
ce
r
class
if
icatio
n
.
B
ased
o
n
th
e
class
if
icatio
n
o
u
tco
m
e,
r
ec
o
m
m
en
d
atio
n
o
n
p
atien
t
tr
ea
tm
en
ts
is
s
u
g
g
ested
.
Fig
u
r
e
1
illu
s
tr
ates
th
e
clo
u
d
-
I
o
T
b
ased
lu
n
g
ca
n
ce
r
class
if
icat
io
n
an
d
r
ec
o
m
m
en
d
atio
n
m
o
d
el.
Fig
u
r
e
1
.
Ov
e
r
all
ar
ch
itectu
r
e
o
f
th
e
lu
n
g
ca
n
ce
r
class
if
icatio
n
an
d
r
ec
o
m
m
e
n
d
atio
n
m
o
d
el
3
.
1
.
Da
t
a
a
cquis
it
io
n
I
n
th
e
d
ata
ac
q
u
is
itio
n
s
tag
e,
C
T
im
ag
es
ar
e
ca
p
tu
r
ed
u
s
in
g
I
o
T
d
ev
ices
an
d
s
to
r
ed
in
clo
u
d
.
T
h
ese
I
o
T
d
ev
ices
ar
e
e
q
u
ip
p
ed
with
s
en
s
o
r
s
ca
p
ab
le
o
f
ca
p
tu
r
in
g
d
etailed
im
ag
es
o
f
t
h
e
lu
n
g
s
,
p
r
o
v
id
in
g
v
alu
ab
l
e
d
ata
f
o
r
an
aly
s
is
.
T
h
e
ca
p
tu
r
ed
C
T
im
ag
es
s
er
v
e
as
in
p
u
t
f
o
r
a
DL
-
b
ased
clo
u
d
-
I
o
T
d
ec
is
io
n
s
u
p
p
o
r
t
s
y
s
tem
s
p
ec
if
ically
d
esig
n
ed
f
o
r
lu
n
g
ca
n
ce
r
class
if
icatio
n
.
T
h
e
p
r
o
ce
s
s
in
clu
d
es
p
r
ep
r
o
ce
s
s
in
g
th
e
C
T
im
ag
es
to
en
h
an
ce
t
h
eir
q
u
al
ity
an
d
r
em
o
v
e
n
o
is
e,
en
s
u
r
i
n
g
ac
cu
r
ate
class
if
icatio
n
r
es
u
lts
.
E
ac
h
ca
p
tu
r
ed
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:
2088
-
8
7
0
8
Hyb
r
id
d
ee
p
lea
r
n
in
g
(
I
LeS
-
N
et)
fo
r
lu
n
g
ca
n
ce
r
cla
s
s
ifica
tio
n
in
…
(
A
ffr
o
s
e
)
1591
im
ag
e
is
r
e
p
r
esen
ted
as
p
ar
t
o
f
th
e
in
p
u
t
C
T
im
a
g
es
d
e
n
o
t
ed
as
=
(
1
,
2
,
.
.
.
,
)
,
wh
er
e
r
ep
r
esen
ts
th
e
to
tal
n
u
m
b
er
o
f
ca
p
tu
r
e
d
im
ag
es.
T
h
r
o
u
g
h
th
is
co
m
p
r
e
h
en
s
iv
e
d
ata
ac
q
u
is
itio
n
p
r
o
ce
s
s
,
th
e
s
y
s
tem
aim
s
to
u
tili
ze
th
e
wea
lth
o
f
in
f
o
r
m
atio
n
g
ath
er
ed
f
r
o
m
I
o
T
d
ev
ic
es to
im
p
r
o
v
e
ea
r
ly
d
etec
tio
n
an
d
class
if
icatio
n
o
f
lu
n
g
ca
n
ce
r
,
ev
en
tu
ally
co
n
tr
ib
u
tin
g
t
o
m
o
r
e
e
f
f
icien
t
p
atie
n
t
ca
r
e
a
n
d
t
r
ea
tm
en
t
o
u
tco
m
e
s
.
Acc
o
r
d
in
g
to
th
is
wo
r
k
,
we
ar
e
c
o
n
s
id
er
in
g
b
en
c
h
m
ar
k
d
ataset
to
g
et
th
e
C
T
i
m
ag
es.
3
.
2
.
G
a
us
s
ia
n f
ilte
ring
t
ec
hn
iqu
e
-
ba
s
ed
pre
-
pro
ce
s
s
i
ng
I
n
th
is
s
tag
e
f
o
llo
win
g
d
ata
ac
q
u
is
itio
n
o
f
C
T
im
ag
es
,
s
ev
er
al
k
ey
s
tep
s
ar
e
u
n
d
er
tak
en
to
en
s
u
r
e
th
at
th
e
r
aw
d
ata
o
b
tain
ed
f
r
o
m
th
e
im
ag
in
g
eq
u
ip
m
en
t
is
o
p
tim
ized
f
o
r
s
u
b
s
eq
u
e
n
t
an
aly
s
is
.
I
n
itially
,
th
e
ac
q
u
ir
ed
C
T
im
ag
es
m
ay
s
u
f
f
er
f
r
o
m
v
ar
io
u
s
a
r
tifa
cts
an
d
n
o
is
e,
wh
ich
ca
n
o
b
s
cu
r
e
im
p
o
r
tan
t
f
ea
tu
r
es
r
elev
an
t
to
lu
n
g
ca
n
ce
r
d
etec
t
io
n
.
So
,
p
r
e
-
p
r
o
ce
s
s
in
g
th
is
i
m
ag
e
is
es
s
en
tial.
T
h
er
ef
o
r
e,
we
h
av
e
u
tili
ze
d
th
e
n
o
is
e
r
ed
u
ctio
n
tech
n
iq
u
e
lik
e
Gau
s
s
ian
f
ilter
in
g
to
en
h
an
ce
C
T
im
ag
e
q
u
ality
wh
ile
p
r
eser
v
in
g
cr
itical
an
ato
m
ical
d
etails.
Mo
r
eo
v
er
,
Gau
s
s
ian
f
ilter
in
g
i
s
wid
ely
em
p
lo
y
e
d
in
im
a
g
e
p
r
o
ce
s
s
in
g
to
d
im
i
n
is
h
n
o
is
e
an
d
ac
h
iev
e
s
m
o
o
th
in
g
ef
f
ec
ts
.
W
h
en
ap
p
l
ied
to
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e,
th
e
Gau
s
s
ian
f
ilter
in
g
en
h
an
ce
s
th
e
in
p
u
t
C
T
im
ag
e
q
u
ality
b
y
m
i
n
im
izin
g
n
o
is
e
wh
ile
r
etain
in
g
cr
u
cial
f
ea
tu
r
es
ess
en
tial
f
o
r
ac
cu
r
ate
clas
s
if
icatio
n
.
T
h
is
tech
n
iq
u
e
in
v
o
l
v
es
th
e
c
o
n
v
o
lu
tio
n
o
f
th
e
im
ag
e
with
a
G
au
s
s
ian
k
er
n
ela
b
ell
-
s
h
ap
ed
c
u
r
v
e
r
ef
lectin
g
th
e
v
alu
e
d
is
tr
ib
u
tio
n
.
E
s
s
en
tially
,
Gau
s
s
ian
f
ilter
in
g
p
er
f
o
r
m
s
a
lin
ea
r
f
ilter
in
g
o
p
e
r
atio
n
ai
m
ed
at
b
lu
r
r
in
g
o
r
s
m
o
o
th
in
g
th
e
im
a
g
e.
B
y
b
en
ef
it
o
f
its
co
n
v
o
l
u
tio
n
p
r
o
ce
s
s
with
th
e
Gau
s
s
ian
k
er
n
el,
th
is
tech
n
iq
u
e
ef
f
ec
tiv
ely
r
e
d
u
ce
s
h
i
g
h
-
f
r
eq
u
en
cy
n
o
is
e
in
th
e
C
T
im
ag
e
wh
ile
m
ain
tain
i
n
g
t
h
e
im
p
o
r
tan
t
f
ea
tu
r
es,
cr
u
cial
f
o
r
s
u
b
s
eq
u
en
t c
lass
if
icatio
n
task
s
.
E
q
u
atio
n
(
1
)
e
x
p
l
ain
th
e
g
au
s
s
ian
k
er
n
el
[
3
3
]
.
(
,
)
=
1
2
2
−
2
+
2
2
2
(
1
)
E
q
u
atio
n
(
1
)
,
th
e
co
o
r
d
in
ates
(
,
)
ar
e
r
elativ
e
to
th
e
ce
n
ter
o
f
th
e
k
er
n
el,
(
,
)
d
en
o
tes
th
e
g
au
s
s
ian
f
u
n
ctio
n
v
al
u
e
at
co
o
r
d
in
ates
(
,
)
,
d
eter
m
in
es th
e
wid
th
o
f
th
e
k
er
n
el
b
y
s
er
v
i
n
g
as th
e
s
tan
d
ar
d
d
ev
iatio
n
o
f
th
e
Gau
s
s
ian
d
is
tr
ib
u
tio
n
.
He
n
ce
,
th
e
o
u
tp
u
t
ac
h
iev
ed
f
r
o
m
th
is
p
r
e
-
p
r
o
ce
s
s
in
g
s
tag
e
is
d
en
o
te
d
as
.
3
.
3
.
I
m
pro
v
ed
B
I
RCH
ba
s
e
d seg
m
ent
a
t
io
n
Af
ter
p
r
ep
r
o
ce
s
s
in
g
s
tag
e,
th
e
s
eg
m
en
tatio
n
p
r
o
ce
d
u
r
e
f
o
cu
s
es
o
n
s
eg
m
en
tin
g
t
h
e
d
is
ea
s
ed
r
eg
io
n
s
f
r
o
m
lu
n
g
ca
n
ce
r
im
ag
es.
T
h
is
alg
o
r
ith
m
o
f
f
er
s
en
h
an
ce
d
ef
f
icien
cy
in
s
h
ap
e
an
aly
s
is
b
y
im
p
r
o
v
i
n
g
th
e
d
is
tan
ce
f
u
n
ctio
n
u
s
ed
f
o
r
e
v
alu
atin
g
p
ix
el
d
is
tan
ce
s
,
th
er
eb
y
e
n
ab
lin
g
m
o
r
e
ac
c
u
r
at
e
s
eg
m
en
tatio
n
o
f
d
is
ea
s
ed
r
eg
io
n
s
.
Sp
ec
if
ically
,
th
e
I
-
B
I
R
C
H
alg
o
r
ith
m
co
n
s
i
s
ts
o
f
f
o
u
r
s
tag
es,
with
im
p
r
o
v
em
en
t
m
a
d
e
in
two
s
tag
es
to
en
s
u
r
e
its
ef
f
ec
tiv
e
n
ess
f
o
r
th
e
lu
n
g
ca
n
ce
r
class
if
icatio
n
.
A
d
etailed
d
ep
ictio
n
o
f
th
e
p
r
o
p
o
s
ed
I
-
B
I
R
C
H
alg
o
r
ith
m
’
s
f
lo
w
ca
n
b
e
estab
lis
h
ed
in
Fig
u
r
e
2
,
p
r
o
v
id
i
n
g
a
co
n
cise
y
et
co
m
p
r
eh
en
s
iv
e
o
v
e
r
v
iew
o
f
its
o
p
er
atio
n
an
d
s
ig
n
if
ican
ce
with
in
th
e
s
eg
m
en
tatio
n
p
r
o
ce
d
u
r
e
f
o
r
l
u
n
g
ca
n
ce
r
C
T
i
m
ag
es.
Fig
u
r
e
2
.
Flo
wch
ar
t
o
f
I
-
B
I
R
C
H
b
ased
s
eg
m
en
tatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
3
,
J
u
n
e
20
2
6
:
1
5
8
8
-
1
6
0
7
1592
a.
Pre
-
p
r
o
ce
s
s
in
g
th
e
im
ag
e
:
B
y
th
e
way
,
th
e
p
r
e
-
p
r
o
ce
s
s
ed
im
ag
e
u
n
d
er
g
o
es
ad
d
itio
n
al
p
r
e
-
p
r
o
ce
s
s
in
g
with
in
th
e
I
-
B
I
R
C
H
alg
o
r
ith
m
.
Key
f
ea
tu
r
es
p
er
tin
e
n
t
to
class
if
icat
io
n
ar
e
s
elec
t
ed
,
f
itted
in
to
co
r
r
esp
o
n
d
in
g
cl
u
s
ter
lab
els an
d
an
y
id
en
tifie
d
o
u
tlier
s
ar
e
e
lim
in
ated
v
ia
f
ea
tu
r
e
r
escalin
g
.
b.
Au
to
m
ated
in
itializatio
n
o
f
th
r
esh
o
ld
:
I
n
th
e
co
n
v
en
tio
n
al
th
r
esh
o
ld
in
itializatio
n
m
eth
o
d
,
d
ata
p
o
in
ts
th
at
ex
ce
ed
th
e
th
r
esh
o
ld
ar
e
a
d
ju
s
ted
to
th
e
th
r
esh
o
ld
v
alu
e.
T
h
is
ad
ju
s
tm
en
t
is
ac
h
iev
ed
b
y
ex
p
an
d
in
g
th
e
s
ca
le
o
f
th
e
leaf
r
a
d
iu
s
,
ai
m
in
g
to
d
ec
r
ea
s
e
t
h
e
s
p
lit
p
ar
en
t
in
th
e
p
r
o
p
o
s
ed
b
ir
c
h
m
o
d
el
[
3
4
]
.
C
o
n
s
eq
u
en
tly
,
t
h
e
au
t
o
m
atic
i
n
itializatio
n
o
f
th
e
t
h
r
esh
o
ld
a
ch
iev
es
th
r
o
u
g
h
an
en
h
an
ce
d
t
h
r
esh
o
ld
,
wh
ich
is
em
p
lo
y
ed
to
ass
ess
u
n
ce
r
tai
n
ties
,
as d
escr
ib
ed
in
(
2
)
.
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(
)
1
22
21
1
1
2
2
2
1
21
2
l
o
g
.
/
l
o
g
.
/
l
o
g
A
A
F
mF
C
F
F
mF
C
C
T
C
T
b
F
Imp
v
m
F
e
v
T
h
r
e
s
h
o
ld
m
F
e
p
p
−
−
−
−
−
−
+
−
=
−
−
(
2
)
Fro
m
(
2
)
,
1
=
∑
=
0
2
=
∑
−
1
=
+
1
,
1
=
1
1
∑
=
0
,
2
=
1
2
∑
−
1
=
0
an
d
1
=
∑
−
1
=
0
,
d
ef
in
es
th
e
im
ag
e
o
v
er
all
g
r
a
y
v
alu
e,
ex
p
lain
s
th
e
av
er
ag
e
v
al
u
e
o
f
g
r
ay
lev
el
o
f
th
e
p
ix
el,
=
2
,
d
en
o
t
es th
e
f
o
ca
l
elem
en
t
o
f
,
m
ass
f
u
n
ctio
n
is
in
d
icate
d
b
y
,
|
|
ca
r
d
in
ality
o
f
,
an
d
|
|
ca
r
d
in
ality
o
f
b
elief
o
f
en
tr
o
p
y
.
c.
B
ir
ch
s
u
b
clu
s
ter
in
g
:
Du
r
i
n
g
th
e
s
u
b
-
clu
s
ter
in
g
s
tag
e,
ea
ch
p
atien
t’
s
d
ata
p
o
in
t,
r
ep
r
e
s
en
tin
g
v
ar
i
o
u
s
f
ea
tu
r
es
s
u
ch
as
tu
m
o
r
s
ize,
lo
ca
tio
n
,
an
d
h
is
to
p
ath
o
lo
g
ica
l
ch
ar
ac
ter
is
tics
,
is
a
s
s
ig
n
ed
t
o
a
s
u
b
-
clu
s
ter
b
ased
o
n
its
s
im
ilar
ity
to
th
e
ce
n
tr
o
id
o
f
th
at
s
u
b
-
clu
s
ter
.
T
h
is
p
r
o
x
im
ity
is
ty
p
ically
m
ea
s
u
r
ed
u
s
in
g
d
is
tan
ce
m
etr
ics
lik
e
E
u
clid
ea
n
d
is
tan
ce
o
r
co
s
in
e
s
im
ilar
ity
.
B
y
iter
ativ
el
y
ass
ig
n
in
g
d
a
ta
p
o
in
ts
to
th
e
n
ea
r
est
s
u
b
-
clu
s
ter
ce
n
tr
o
id
,
t
h
e
alg
o
r
ith
m
g
r
ad
u
ally
r
e
f
in
es
th
e
s
eg
m
e
n
tatio
n
,
e
f
f
ec
tiv
ely
b
r
ea
k
i
n
g
d
o
wn
lar
g
er
,
h
eter
o
g
en
e
o
u
s
clu
s
ter
s
o
f
lu
n
g
ca
n
ce
r
ca
s
es in
to
s
m
aller
,
m
o
r
e
h
o
m
o
g
en
e
o
u
s
g
r
o
u
p
s
.
d.
Alter
in
g
d
is
tan
ce
m
etr
ices
in
th
e
b
aselin
e
b
ir
ch
:
C
o
n
v
en
tio
n
ally
,
th
e
d
is
tan
ce
[
3
5
]
is
ca
lcu
lated
as
g
iv
en
in
(
3
)
.
=
∑
(
|
−
|
,
−
|
−
|
)
=
1
(
3
)
W
h
ile
ef
f
ec
tiv
e
f
o
r
lar
g
e
d
ata
s
ets,
th
is
m
eth
o
d
m
ay
f
ac
e
c
h
allen
g
es
with
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
an
d
s
ca
lab
ilit
y
wh
en
d
ea
lin
g
with
ex
tr
em
ely
lar
g
e
o
r
h
i
g
h
-
d
im
e
n
s
io
n
al
d
ata.
T
h
e
r
ef
o
r
e,
th
e
d
is
tan
ce
is
im
p
r
o
v
e
d
b
y
co
m
b
i
n
in
g
th
e
I
m
p
r
o
v
e
d
L
ee
d
is
tan
ce
an
d
E
u
clid
ea
n
d
is
tan
ce
as e
x
p
r
ess
ed
i
n
(
4
)
.
=
{
∑
(
|
−
|
,
−
|
−
|
)
=
1
(
√
∑
(
)
2
,
=
1
√
∑
(
)
2
=
1
)
+
(
|
−
|
)
}
+
√
∑
(
−
)
2
=
1
/
2
(
4
)
Eq
u
atio
n
(
4
)
,
,
r
ep
r
esen
t
p
ix
el
s
th
at
ar
e
ad
jace
n
t
to
o
n
e
an
o
th
er
.
T
h
is
im
p
r
o
v
em
en
t
o
p
tim
izes
clu
s
ter
in
g
ac
cu
r
ac
y
,
m
e
m
o
r
y
u
s
ag
e,
a
n
d
tim
e
co
m
p
lex
ity
b
y
ef
f
icien
tly
s
ca
lin
g
to
lar
g
e
d
atasets
.
T
h
e
f
in
al
s
tep
in
I
-
B
I
R
C
H
in
v
o
lv
es
o
u
tp
u
ttin
g
th
e
clu
s
ter
s
,
ty
p
ically
d
en
o
ted
as
clu
s
ter
1
,
clu
s
ter
2
.
T
h
ese
clu
s
ter
s
r
ep
r
esen
t
th
e
f
in
al
s
eg
m
en
tatio
n
o
f
t
h
e
d
at
aset
b
ased
o
n
th
e
clu
s
ter
in
g
p
er
f
o
r
m
ed
b
y
th
e
I
-
B
I
R
C
H.
Af
ter
o
b
tain
in
g
th
e
clu
s
ter
s
,
th
e
s
eg
m
en
ted
im
ag
e
ca
n
b
e
g
e
n
er
ated
b
y
ass
ig
n
in
g
ea
ch
p
ix
el
in
t
h
e
o
r
ig
in
al
im
ag
e
to
th
e
cl
u
s
ter
it
b
elo
n
g
s
to
.
E
ac
h
cl
u
s
ter
r
ep
r
esen
ts
a
d
is
tin
ct
r
eg
io
n
o
r
g
r
o
u
p
o
f
p
ix
els
in
th
e
im
ag
e
th
at
s
h
ar
e
s
im
ilar
ch
ar
a
cter
is
tics
.
T
h
u
s
,
th
e
o
u
tp
u
t
ac
h
iev
ed
f
r
o
m
th
is
I
-
B
I
R
C
H
b
ased
s
eg
m
en
tatio
n
is
im
p
lied
as
.
3
.
4
.
F
e
a
t
ure
ex
t
r
a
ct
io
n
Af
ter
s
eg
m
en
tatio
n
,
f
r
o
m
th
e
s
eg
m
en
ted
im
ag
e
,
r
elev
an
t
f
ea
tu
r
es
ar
e
ex
tr
ac
ted
.
T
h
is
p
r
o
ce
s
s
aim
s
to
ca
p
tu
r
e
d
is
tin
ctiv
e
f
ea
tu
r
es
o
f
th
e
s
eg
m
en
ted
r
e
g
io
n
s
,
wh
ich
aid
s
in
d
is
tin
g
u
is
h
in
g
h
ea
lth
y
an
d
ca
n
ce
r
o
u
s
tis
s
u
e.
T
y
p
ically
,
f
ea
tu
r
es
lik
e
s
h
ap
e
f
ea
tu
r
es,
I
m
p
r
o
v
e
d
L
GXP
b
ased
f
ea
tu
r
e
,
an
d
co
lo
r
f
ea
tu
r
es
ar
e
r
etr
iev
ed
.
E
ac
h
f
ea
tu
r
e
ca
p
tu
r
es
d
iv
er
s
e
in
f
o
r
m
atio
n
ab
o
u
t
th
e
lu
n
g
n
o
d
u
les.
T
h
er
ef
o
r
e,
ex
tr
ac
tin
g
s
h
ap
e
,
I
-
L
GXP,
an
d
co
lo
r
f
ea
tu
r
es
al
lo
ws
th
e
class
if
ier
to
u
n
d
er
s
tan
d
wh
at
p
atter
n
a
n
o
d
u
le
h
as
an
d
h
o
w
th
e
n
o
d
u
le
lo
o
k
s
an
d
wh
at
d
en
s
ity
it
h
as
.
W
ith
th
ese
d
etails,
th
e
cla
s
s
if
icatio
n
is
co
n
d
u
cted
.
E
x
tr
ac
ted
ea
ch
f
ea
tu
r
e
is
d
etaile
d
.
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:
2088
-
8
7
0
8
Hyb
r
id
d
ee
p
lea
r
n
in
g
(
I
LeS
-
N
et)
fo
r
lu
n
g
ca
n
ce
r
cla
s
s
ifica
tio
n
in
…
(
A
ffr
o
s
e
)
1593
3
.
4
.
1
.
Sh
a
pe
f
ea
t
ures
Sh
ap
e
f
ea
tu
r
es
p
lay
a
v
ital
r
o
le
in
ch
a
r
ac
ter
izin
g
t
h
e
g
e
o
m
et
r
ic
p
r
o
p
er
ties
o
f
th
e
s
eg
m
e
n
te
d
r
eg
io
n
s
b
y
in
p
u
ttin
g
th
e
s
eg
m
en
ted
i
m
ag
e
.
I
t
r
ef
er
s
to
ch
ar
ac
ter
is
tics
o
r
d
escr
ip
t
o
r
s
d
e
r
iv
ed
f
r
o
m
th
e
s
h
ap
es
o
f
r
eg
io
n
s
with
in
th
e
s
eg
m
en
ted
lu
n
g
im
ag
es.
Her
e,
ce
r
tain
g
eo
m
etr
ic
ch
ar
ac
ter
is
tics
s
u
ch
as
ar
ea
,
m
o
m
en
t,
p
er
im
eter
,
ep
s
ilo
n
,
an
d
c
o
n
v
e
x
ity
ar
e
b
ein
g
r
etr
iev
ed
as th
e
s
h
ap
e
f
ea
tu
r
e.
−
Ar
ea
:
I
t
q
u
an
tifie
s
th
e
to
tal
n
u
m
b
er
o
f
p
ix
els
with
in
ea
c
h
s
eg
m
en
ted
r
e
g
io
n
[
3
6
]
an
d
it
s
ex
p
r
ess
io
n
is
p
r
o
v
id
e
d
in
(
5
)
.
=
∑
∑
(
,
)
=
1
=
1
(
5
)
h
er
e,
im
ag
e
s
ize
is
d
en
o
ted
as
,
,
s
p
atial
co
o
r
d
in
ates a
r
e
r
ep
r
es
en
ted
as
,
.
−
Mo
m
en
t:
Mo
m
e
n
ts
ar
e
m
at
h
e
m
atica
l
d
escr
ip
to
r
s
th
at
ca
p
tu
r
e
d
if
f
e
r
en
t
asp
ec
ts
o
f
th
e
s
h
ap
e
an
d
s
p
atial
d
is
tr
ib
u
tio
n
o
f
p
ix
els with
in
a
s
eg
m
en
ted
ar
ea
.
−
Per
im
eter
:
T
h
e
p
er
im
eter
f
ea
tu
r
e
m
ea
s
u
r
es
th
e
to
tal
len
g
t
h
o
f
th
e
b
o
u
n
d
ar
y
o
r
o
u
tlin
e
o
f
th
e
s
eg
m
en
ted
r
eg
io
n
[
3
6
]
as d
ef
in
ed
b
y
(
6
)
.
=
∑
√
(
1
−
−
1
)
+
(
1
−
−
1
)
2
=
1
(
6
)
h
er
e,
th
e
-
th
p
ix
el
o
f
th
e
s
p
atia
l c
o
o
r
d
in
ates is r
ep
r
esen
te
d
as
,
.
−
E
p
s
ilo
n
:
E
p
s
ilo
n
is
a
p
ar
am
eter
u
s
ed
in
s
h
ap
e
an
aly
s
is
t
o
d
ef
in
e
th
e
m
ax
im
u
m
allo
wab
le
d
is
tan
ce
b
etwe
en
a
co
n
t
o
u
r
a
n
d
its
ap
p
r
o
x
im
atio
n
.
−
C
o
n
v
ex
ity
: Co
n
v
e
x
ity
m
ea
s
u
r
es th
e
d
eg
r
ee
to
wh
ich
a
s
eg
m
en
ted
r
eg
io
n
is
co
n
v
ex
o
r
co
n
c
av
e
in
s
h
ap
e.
T
h
u
s
,
th
e
o
u
tp
u
t a
ttain
ed
f
r
o
m
th
is
s
h
ap
e
f
ea
tu
r
e
is
s
p
ec
if
ied
as
ℎ
.
3
.
4
.
2
.
I
m
pro
v
ed
L
G
XP
ba
s
ed
f
ea
t
ure
L
GXP
is
a
f
ea
tu
r
e
d
escr
ip
to
r
th
at
f
o
cu
s
es
o
n
en
co
d
in
g
lo
ca
l
g
r
ad
ien
t
in
f
o
r
m
atio
n
a
n
d
in
c
o
r
p
o
r
ati
n
g
XOR
p
atter
n
s
to
r
ep
r
esen
t
tex
tu
r
e
v
ar
iatio
n
s
with
in
th
e
s
eg
m
en
ted
r
eg
io
n
s
.
Du
r
i
n
g
th
is
p
r
o
ce
s
s
,
th
e
s
eg
m
en
ted
im
ag
es
ar
e
an
aly
z
ed
at
a
lo
ca
l
lev
el
to
c
o
m
p
u
te
g
r
ad
ien
t
in
f
o
r
m
atio
n
,
wh
ich
p
r
o
v
id
es
in
s
ig
h
ts
in
to
th
e
in
ten
s
ity
ch
an
g
es
ac
r
o
s
s
n
eig
h
b
o
r
in
g
p
ix
els.
T
h
is
co
n
ce
p
t
en
co
m
p
ass
es
f
iv
e
p
r
im
ar
y
s
tag
es
co
n
tain
s
im
ag
e
ac
q
u
is
itio
n
,
n
o
is
e
s
u
p
p
r
ess
io
n
,
im
p
r
o
v
ed
g
r
ad
ie
n
t
co
m
p
u
tatio
n
,
L
GXOR,
an
d
h
is
to
g
r
am
o
f
im
a
g
e
[
3
7
]
.
T
h
er
eb
y
,
th
e
f
lo
wc
h
ar
t illu
s
tr
atin
g
th
e
p
r
o
ce
d
u
r
e
o
f
th
e
I
L
G
XP is d
ep
icted
in
Fig
u
r
e
3
.
Fig
u
r
e
3
.
Flo
wch
ar
t
o
f
I
m
p
r
o
v
ed
L
GXP
a.
I
m
ag
e
ac
q
u
is
itio
n
T
h
e
in
p
u
t
s
eg
m
en
ted
im
a
g
e
is
co
n
v
er
ted
i
n
to
a
g
r
ay
s
ca
le
i
m
ag
e
(
n
o
n
-
c
o
lo
r
ed
)
,
s
im
p
lify
i
n
g
th
e
an
aly
s
is
p
r
o
ce
s
s
.
T
h
e
in
ten
s
ity
o
f
th
is
g
r
ay
s
ca
le
im
ag
e
r
an
g
e
s
f
r
o
m
0
to
2
5
5
.
T
h
er
ef
o
r
e,
th
e
ex
p
r
ess
io
n
f
o
r
th
e
g
r
ay
s
ca
le
im
ag
e
is
elu
cid
ated
as f
o
llo
ws:
An
im
ag
e
is
co
m
p
r
is
ed
o
f
p
i
x
els
o
r
g
an
ized
in
r
o
ws
an
d
c
o
lu
m
n
s
to
d
escr
ib
e
its
d
im
en
s
io
n
s
.
T
h
e
wid
th
an
d
h
ei
g
h
t
o
f
ea
ch
co
l
u
m
n
a
n
d
r
o
w
in
an
im
ag
e
ar
e
d
eter
m
in
e
d
b
y
its
r
eso
lu
tio
n
.
T
h
e
im
a
g
e
to
tal
n
u
m
b
er
o
f
p
ix
els
is
m
ea
n
t
as
th
at
s
ig
n
if
ies
its
d
im
en
s
io
n
s
.
Pix
els
in
ten
s
ity
m
atr
ix
p
o
s
itio
n
ed
at
co
o
r
d
in
ates
(
,
)
with
in
th
e
d
ig
it
al
im
ag
e
is
u
tili
ze
d
t
o
r
e
p
r
esen
t
it,
wh
er
e
in
d
icate
s
wid
th
an
d
in
d
icate
s
h
eig
h
t.
T
ak
e
(
=
1
,
2
,
3
,
.
.
,
,
=
1
,
2
,
3
,
.
.
,
)
as
g
r
ay
s
ca
le,
h
er
e
r
ep
r
esen
ts
th
e
d
ig
ital
im
ag
e’
s
in
ten
s
ity
at
t
h
e
s
p
ec
i
f
ied
p
o
s
itio
n
.
.
.
(
,
)
.
T
h
e
im
ag
e
m
ay
b
e
s
u
b
ject
to
f
lu
ct
u
atin
g
n
o
is
e
,
af
f
ec
tin
g
th
e
p
ix
el
at
lo
ca
tio
n
(
,
)
.
C
o
n
s
eq
u
en
tly
,
th
e
co
m
b
in
e
d
s
ettin
g
,
as
ex
p
r
ess
ed
in
(
7
)
,
i
s
em
p
lo
y
ed
to
d
ef
in
e
th
e
im
a
g
e’
s
g
r
a
y
lev
el
i
n
ten
s
ity
.
T
h
u
s
,
th
e
g
r
ay
s
ca
le
i
m
ag
e
ca
n
b
e
r
ep
r
esen
ted
as
(
,
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
3
,
J
u
n
e
20
2
6
:
1
5
8
8
-
1
6
0
7
1594
,
=
,
+
,
(
7
)
b.
No
is
e
s
u
p
p
r
ess
io
n
Firstl
y
,
th
e
g
r
ay
s
ca
le
im
ag
e
(
,
)
is
f
ed
as
in
p
u
t
to
th
e
Gab
o
r
f
ilter
.
T
h
is
co
m
b
in
atio
n
h
as
estab
lis
h
ed
o
p
tim
is
tic
o
u
tco
m
es.
Gab
o
r
f
ilter
c
h
ar
ac
ter
iz
ed
as
b
a
n
d
p
ass
f
ilter
s
,
th
at
s
elec
tiv
ely
allo
w
a
s
p
ec
if
ied
r
an
g
e
o
f
f
r
eq
u
en
ci
es
to
p
ass
th
r
o
u
g
h
.
Ma
th
e
m
atica
lly
,
a
1
D
Gab
o
r
f
ilte
r
is
r
ep
r
esen
ted
as
co
n
v
o
l
u
tio
n
o
f
a
f
u
n
ctio
n
s
in
u
s
o
id
al,
as
d
escr
ib
ed
in
(
8
)
,
with
a
Gau
s
s
ian
f
u
n
ctio
n
.
Her
e,
an
d
d
en
o
te
th
e
s
in
u
s
o
id
al
wav
e’
s
f
r
eq
u
e
n
cy
an
d
p
h
ase
o
f
f
s
ets,
wh
ile
an
d
r
ep
r
esen
t
th
e
wa
v
elen
g
th
an
d
s
tan
d
a
r
d
d
ev
iatio
n
alo
n
g
th
e
x
-
ax
is
o
f
th
e
Gab
o
r
f
u
n
ctio
n
.
(
)
=
(
−
2
2
2
)
(
(
2
+
)
)
=
+
,
(
=
√
−
1
)
(
8
)
Giv
en
th
at
th
e
im
ag
e
s
ig
n
if
ies
a
2
D
f
u
n
ctio
n
,
eq
u
atio
n
(
8
)
h
as
b
ee
n
ad
ju
s
ted
to
(
9
)
,
wh
e
r
e
th
e
r
ea
l
an
d
im
ag
i
n
ar
y
p
ar
ts
ar
e
d
en
o
ted
as
(
,
)
=
(
−
2
+
2
2
2
2
)
(
2
+
)
an
d
(
,
)
=
(
−
2
+
2
2
2
2
)
(
2
+
)
.
Her
e
,
=
+
;
=
−
+
.
Als
o
,
th
e
c
o
n
v
e
n
t
io
n
al
r
ea
l p
ar
t is g
iv
en
i
n
(
1
0
)
.
(
,
)
=
(
−
2
+
2
2
2
2
)
(
(
2
+
)
)
(
9
)
(
,
;
,
,
,
)
=
(
−
2
+
2
2
2
2
)
(
2
+
)
(
1
0
)
Ho
wev
er
,
e
x
ce
s
s
iv
e
p
ar
am
ete
r
ad
ju
s
tm
en
ts
m
ay
lead
t
o
o
v
e
r
f
itti
n
g
,
ca
u
s
in
g
t
h
e
f
ilter
to
b
ec
o
m
e
to
o
s
p
ec
ialized
an
d
p
o
te
n
tially
less
ef
f
ec
tiv
e
at
r
ec
o
g
n
izin
g
b
r
o
ad
er
p
atter
n
s
o
r
v
ar
iatio
n
s
in
tex
tu
r
es.
T
h
u
s
,
th
e
r
ea
l p
ar
t o
f
th
e
g
r
ay
s
ca
le
is
en
h
an
ce
d
in
th
is
I
-
L
GXP
[
3
8
]
as p
r
o
v
i
d
ed
in
(
1
1
)
.
(
,
;
,
,
,
)
=
(
−
2
+
2
2
2
2
)
(
2
+
)
.
[
2
,
(
,
)
(
2
′
)
]
(
1
1
)
In
(
1
1
)
,
th
e
(
,
)
=
∑
2
(
−
)
2
=
1
∑
(
+
)
=
1
+
∗
100
,
=
+
,
=
−
+
,
(
,
)
=
[
∑
(
−
̄
)
(
−
̄
)
=
1
√
∑
(
−
̄
)
2
∑
(
−
̄
)
2
=
1
=
1
]
,
̄
,
̄
d
en
o
tes
th
e
av
er
ag
e
v
alu
e.
B
y
ad
ju
s
tin
g
th
e
p
ar
am
eter
s
en
h
an
ce
s
th
e
f
ilter
’
s
a
b
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
v
a
r
io
u
s
te
x
tu
r
e
p
att
er
n
s
in
im
ag
es,
im
p
r
o
v
in
g
o
v
er
all
im
ag
e
q
u
ality
an
d
an
aly
s
is
ac
cu
r
ac
y
.
Ma
th
em
atica
lly
th
e
g
r
ay
s
ca
le
im
ag
e
is
r
ep
r
esen
ted
b
y
co
m
b
in
in
g
it
with
th
e
Gab
o
r
f
ilter
t
o
f
ilter
o
u
t
u
n
d
esira
b
le
d
ata.
T
h
is
f
ilter
in
g
p
r
o
ce
s
s
aim
s
to
b
lu
r
th
e
im
ag
e.
C
o
n
s
id
er
in
g
im
a
g
es
in
p
u
t
an
d
o
u
t
p
u
t
o
f
t
h
e
f
ilter
ar
e
d
en
o
ted
as
(
,
)
an
d
′
(
,
)
,
Fig
u
r
e
4
illu
s
tr
ates a
p
o
ten
tial sm
o
o
th
in
g
ar
r
an
g
em
en
t.
c.
Gr
ad
ien
t c
o
m
p
u
tatio
n
T
h
e
m
ain
p
r
o
ce
s
s
is
o
n
co
m
p
u
tin
g
g
r
ad
ie
n
ts
an
d
p
ix
el
p
o
in
ts
id
en
tific
atio
n
with
th
e
h
ig
h
est
ch
an
g
e
in
g
r
ay
s
ca
le
in
ten
s
ity
,
also
t
h
e
So
b
el
o
p
e
r
ato
r
b
ein
g
u
s
ed
to
co
m
p
u
te
g
r
ad
ien
ts
.
T
h
is
o
p
er
a
to
r
o
p
e
r
ates
in
2
D,
s
p
ec
if
ically
in
th
e
an
d
d
ir
ec
tio
n
s
.
E
q
u
atio
n
(
1
2
)
r
ep
r
esen
ts
th
e
m
atr
ices
f
o
r
t
h
e
So
b
el
o
p
er
ato
r
s
in
th
ese
two
d
ir
ec
tio
n
s
ar
e
an
d
.
=
[
−
1
0
+
1
−
2
0
+
2
−
1
0
+
1
]
=
[
−
1
+
2
+
1
0
0
0
−
1
−
2
−
1
]
(
1
2
)
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:
2088
-
8
7
0
8
Hyb
r
id
d
ee
p
lea
r
n
in
g
(
I
LeS
-
N
et)
fo
r
lu
n
g
ca
n
ce
r
cla
s
s
ifica
tio
n
in
…
(
A
ffr
o
s
e
)
1595
I
n
(
1
2
)
,
an
d
s
p
ec
if
ies
th
e
m
a
tr
ices
wh
ich
ar
e
co
n
v
o
l
v
ed
with
th
e
b
lu
r
r
ed
im
ag
e
p
ix
els
(
,
)
in
th
e
an
d
d
ir
ec
tio
n
s
,
r
esp
ec
tiv
ely
.
T
h
e
p
ix
el
p
o
in
t g
r
a
d
ien
t
at
(
,
)
is
g
iv
en
in
(
1
3
)
a
n
d
(
1
4
)
.
=
∗
(
,
)
(
1
3
)
=
∗
(
,
)
(
1
4
)
Eq
u
atio
n
(
1
5
)
ca
lcu
lates th
e
m
ag
n
itu
d
e
o
f
th
e
e
d
g
e,
th
at
is
d
e
n
o
ted
as
.
=
√
2
+
2
(
1
5
)
T
h
e
ar
ctan
o
f
th
e
r
atio
b
etwe
e
n
an
d
p
r
o
v
id
es th
e
c
o
m
p
u
te
d
g
r
ad
ien
t
d
ir
ec
tio
n
s
at
th
e
p
ix
e
l p
o
in
t
(
,
)
.
Fig
u
r
e
4
.
I
m
ag
e
is
s
m
o
o
th
e
d
t
h
r
o
u
g
h
n
o
is
e
s
u
p
p
r
ess
io
n
u
s
in
g
a
m
o
d
if
ied
Gab
o
r
f
ilter
d.
L
GXOR
Af
ter
co
m
p
u
tin
g
th
e
g
r
ad
ien
t,
L
GXOR
f
ea
tu
r
es
ar
e
r
etr
iev
e
d
.
Fo
r
lu
n
g
ca
n
ce
r
class
if
icati
o
n
,
Gab
o
r
f
ilter
s
d
em
o
n
s
tr
ate
co
n
s
id
er
a
b
le
ef
f
icien
c
y
a
n
d
ad
eq
u
ac
y
in
ca
p
tu
r
in
g
th
eir
in
d
iv
id
u
al
f
ea
tu
r
es.
B
o
th
b
an
d
p
ass
an
d
Gab
o
r
f
ilter
s
a
r
e
em
p
lo
y
ed
to
e
x
tr
ac
t
f
ea
t
u
r
es.
T
h
is
r
ep
r
esen
tatio
n
i
n
v
o
lv
es
co
n
ca
ten
atin
g
s
ig
n
if
ican
t b
in
ar
y
v
ar
iab
les d
e
r
iv
ed
f
r
o
m
th
e
p
r
im
ar
y
im
ag
e
p
ix
els.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
3
,
J
u
n
e
20
2
6
:
1
5
8
8
-
1
6
0
7
1596
e.
His
to
g
r
am
o
f
im
ag
e
I
-
L
GXP
in
v
o
lv
es
o
b
tain
in
g
th
e
h
is
to
g
r
am
o
f
th
e
im
ag
e
in
th
e
f
in
al
s
tag
e.
T
h
e
h
is
to
g
r
am
o
f
im
ag
e
is
d
er
iv
ed
f
r
o
m
L
GXOR
r
etr
iev
ed
f
ea
tu
r
e’
s
o
u
tp
u
t.
E
q
u
atio
n
(
1
6
)
r
ep
r
esen
ts
a
m
ath
e
m
atica
l
eq
u
atio
n
f
o
r
d
etailin
g
h
o
w
th
ese
f
ea
tu
r
es
a
r
e
p
ar
titi
o
n
ed
in
t
o
s
u
b
-
b
lo
ck
s
.
Her
e,
,
,
s
ig
n
if
ies
th
e
h
is
to
g
r
am
o
f
th
e
ℎ
s
u
b
-
b
lo
ck
o
f
t
h
e
(
=
1
,
2
,
.
.
,
)
L
GXOR
m
ap
.
T
wo
L
GXOR
d
escr
ip
to
r
s
f
o
r
lu
n
g
ca
n
ce
r
,
d
en
o
te
d
as
1
an
d
1
ar
e
e
v
alu
ated
ac
c
o
r
d
in
g
to
(
1
7
)
.
E
q
u
atio
n
(
1
8
)
illu
s
tr
a
tes
th
e
o
p
er
atio
n
o
f
h
is
to
g
r
am
in
ter
s
ec
tio
n
;
th
e
to
tal
am
o
u
n
t o
f
h
is
to
g
r
am
b
in
s
is
in
d
icatin
g
as
ℎ
.
=
[
0
,
0
,
1
,
.
.
.
,
0
,
0
,
;
0
−
1
,
0
−
1
,
,
.
.
,
0
−
1
,
0
−
1
,
]
(
1
6
)
′
(
1
,
2
)
=
∑
∑
∑
∩
(
,
,
1
,
,
,
2
)
=
1
−
1
=
0
−
1
=
0
(
1
7
)
∩
(
1
,
2
)
=
∑
(
ℎ
1
,
ℎ
2
)
ℎ
=
1
(
1
8
)
T
h
u
s
,
th
e
I
-
L
GXP
o
u
tp
u
t a
c
h
iev
ed
f
r
o
m
th
is
s
tag
e
is
im
p
lied
as
−
.
3
.
4
.
3
.
Co
lo
r
f
e
a
t
ures
W
h
en
an
aly
zin
g
s
eg
m
en
ted
C
T
im
ag
es,
co
lo
r
f
ea
tu
r
es
lik
e
co
lo
r
h
is
to
g
r
am
,
m
ea
n
,
m
ed
ian
,
an
d
s
k
ewn
ess
p
r
o
v
id
e
v
alu
ab
le
in
s
ig
h
ts
in
to
th
e
d
is
tr
ib
u
tio
n
a
n
d
ch
a
r
ac
ter
is
tics
o
f
p
ix
el
in
t
en
s
ities
with
in
th
e
s
eg
m
en
ted
r
eg
i
o
n
s
.
Her
e,
th
e
s
eg
m
en
ted
im
a
g
e
is
g
iv
en
as a
n
in
p
u
t f
o
r
r
etr
ie
v
e
th
e
C
o
lo
r
f
e
atu
r
es.
a.
C
o
lo
r
h
is
to
g
r
am
:
T
h
is
f
ea
tu
r
e
r
ep
r
esen
ts
th
e
f
r
eq
u
en
cy
d
is
tr
ib
u
tio
n
o
f
p
ix
el
in
ten
s
ities
with
in
th
e
s
eg
m
en
ted
r
eg
io
n
ac
r
o
s
s
d
if
f
er
en
t
co
lo
r
ch
an
n
els
[
3
9
]
.
I
t
s
h
o
ws
h
o
w
p
r
ev
alen
t
ea
ch
in
ten
s
ity
lev
el
is
,
p
r
o
v
id
i
n
g
an
o
v
e
r
all
v
iew
o
f
t
h
e
co
lo
r
d
is
tr
ib
u
tio
n
as d
escr
i
b
ed
b
y
(
1
9
)
.
=
{
ℎ
[
1
]
,
ℎ
[
12
]
,
.
.
.
.
.
,
ℎ
[
]
=
1
,
0
≤
ℎ
[
]
≤
1
}
(
1
9
)
T
h
e
f
r
e
q
u
en
c
y
o
f
p
ix
els
is
d
en
o
ted
as
ℎ
[
]
,
co
r
r
esp
o
n
d
in
g
to
th
e
-
th
co
lo
r
p
r
esen
t
in
t
h
e
i
m
ag
e
is
d
escr
ib
ed
in
(
2
0
)
.
ℎ
[
]
=
∑
∑
{
1
(
(
,
)
=
)
0
(
ℎ
)
2
−
1
=
0
1
−
1
=
0
1
×
2
(
2
0
)
I
n
(
2
0
)
,
th
e
im
ag
e
wid
th
is
d
e
n
o
ted
b
y
1
an
d
th
e
im
a
g
e
h
eig
h
t is si
g
n
if
ied
b
y
2
.
b.
Me
an
:
I
t is a
m
ea
s
u
r
e
o
f
th
e
a
v
er
ag
e
b
r
ig
h
tn
ess
o
r
c
o
lo
r
v
al
u
e
[
4
0
]
an
d
its
ex
p
r
ess
io
n
is
d
ef
in
ed
in
(
2
1
)
.
=
1
∑
∑
=
1
=
1
(
2
1
)
Fro
m
(
2
1
)
,
an
d
m
ea
n
s
t
h
e
d
im
en
s
io
n
o
f
t
h
e
im
a
g
e
a
n
d
co
l
o
r
v
alu
e
is
d
en
o
te
d
b
y
o
n
co
l
u
m
n
an
d
r
o
w
.
c.
Me
d
ian
: I
t d
en
o
tes th
e
m
ed
ian
v
alu
e
with
in
th
e
s
o
r
ted
lis
t o
f
p
ix
el
in
ten
s
ities
with
in
th
e
s
eg
m
en
ted
ar
ea
.
d.
Sk
ewn
ess
:
I
t
ev
alu
ates
th
e
as
y
m
m
etr
y
o
f
th
e
p
ix
el
in
ten
s
ity
d
is
tr
ib
u
tio
n
with
in
th
e
s
eg
m
en
ted
r
eg
io
n
.
Sk
ewn
ess
ca
n
p
r
o
v
id
e
in
s
ig
h
ts
in
to
th
e
s
h
ap
e
an
d
u
n
if
o
r
m
ity
o
f
th
e
co
lo
r
d
is
tr
ib
u
ti
o
n
[
4
0
]
an
d
its
m
ath
em
atica
l c
alcu
latio
n
is
d
e
s
cr
ib
ed
in
(
2
2
)
.
=
∑
∑
(
)
3
=
1
=
1
(
2
2
)
T
h
u
s
,
th
e
o
u
tp
u
t
g
o
tten
f
r
o
m
t
h
is
s
h
ap
e
f
ea
tu
r
e
is
m
ea
n
t
as,
.
L
astl
y
,
th
e
r
elev
an
t
f
ea
tu
r
es
lik
e
s
h
ap
e
f
ea
tu
r
es
ℎ
,
I
-
L
GXP
−
,
an
d
c
o
lo
r
f
ea
tu
r
es
ar
e
r
etr
iev
e
d
.
T
h
er
ef
o
r
e,
th
e
r
etr
iev
ed
f
ea
tu
r
es a
r
e
in
d
icate
d
as,
.
3
.
5
.
L
ung
ca
ncer
cla
s
s
if
ica
t
io
n us
ing
I
L
eS
-
Net
m
o
del
L
ung
ca
n
ce
r
class
if
icatio
n
is
an
ess
en
tial
s
tag
e
in
th
is
p
r
o
p
o
s
ed
I
L
eS
-
Net
m
o
d
el.
T
h
e
r
etr
iev
ed
f
ea
tu
r
es
,
s
er
v
es
as
in
p
u
t
f
o
r
c
lass
if
y
in
g
lu
n
g
ca
n
ce
r
.
T
o
ac
h
iev
e
th
is
task
,
a
n
o
v
el
h
y
b
r
id
I
L
eS
-
Net
m
o
d
el
is
in
tr
o
d
u
ce
d
a
n
d
u
tili
ze
d
f
o
r
class
if
y
in
g
lu
n
g
ca
n
ce
r
.
T
h
is
h
y
b
r
id
m
o
d
el
c
o
m
b
in
e
s
th
e
I
L
eNe
t
-
5
an
d
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:
2088
-
8
7
0
8
Hyb
r
id
d
ee
p
lea
r
n
in
g
(
I
LeS
-
N
et)
fo
r
lu
n
g
ca
n
ce
r
cla
s
s
ifica
tio
n
in
…
(
A
ffr
o
s
e
)
1597
Sq
u
ee
ze
Net
as
s
h
o
wn
in
Fig
u
r
e
5
p
r
esen
ts
a
co
m
p
r
eh
e
n
s
iv
e
m
o
d
el
f
o
r
class
if
y
in
g
lu
n
g
ca
n
ce
r
.
Su
b
s
eq
u
en
t
s
ec
tio
n
s
will p
r
o
v
id
e
d
etailed
ex
p
lan
atio
n
s
o
f
th
ese
two
class
if
ier
s
.
Fig
u
r
e
5
.
Pro
p
o
s
ed
h
y
b
r
i
d
I
L
e
S
-
Net
m
o
d
el
f
o
r
lu
n
g
ca
n
ce
r
c
lass
if
icatio
n
3
.
5
.
1
.
P
ro
ce
du
ra
l e
x
pla
na
t
io
n o
f
I
L
eNe
t
-
5
m
o
del
C
o
n
v
en
tio
n
ally
,
th
e
L
eNe
t
-
5
[
4
1
]
ar
c
h
itectu
r
e
is
th
e
ef
f
ec
tiv
e
p
r
o
ce
s
s
f
o
r
ca
n
ce
r
class
if
icatio
n
v
ia
C
T
im
ag
es.
T
h
is
L
eNe
t
-
5
m
o
d
el
co
n
tain
s
o
f
two
c
o
n
v
o
lu
ti
o
n
al
lay
er
s
,
two
av
e
r
ag
e
p
o
o
l
in
g
lay
er
s
,
an
d
FC
lay
er
with
r
etr
iev
ed
f
ea
tu
r
es
as in
p
u
t a
s
s
h
o
wn
in
Fig
u
r
e
6
.
Fig
u
r
e
6
.
C
o
n
v
en
tio
n
al
L
eNe
t
-
5
m
o
d
el
T
h
e
co
n
v
o
lu
tio
n
al
lay
er
s
ar
e
r
esp
o
n
s
ib
le
f
o
r
d
etec
tin
g
im
p
o
r
tan
t
p
atter
n
s
an
d
f
ea
tu
r
es
in
d
icativ
e
o
f
lu
n
g
a
b
n
o
r
m
alities
,
s
u
ch
as
tu
m
o
r
m
ass
es
o
r
ir
r
eg
u
lar
ities
in
tis
s
u
e
d
en
s
ity
.
T
h
e
o
u
tp
u
t
f
r
o
m
th
ese
la
y
er
s
is
th
en
f
ed
i
n
to
FC
lay
er
s
,
wh
er
e
th
e
h
ig
h
-
lev
el
f
ea
tu
r
es a
r
e
lea
r
n
ed
an
d
m
ap
p
ed
to
th
e
o
u
tp
u
t
class
es.
I
n
th
is
I
L
eS
-
Net
m
o
d
el,
an
I
L
eNe
t
-
5
ar
ch
itectu
r
e
is
p
r
o
p
o
s
e
d
,
co
m
p
r
is
in
g
s
ev
e
r
al
k
e
y
co
m
p
o
n
e
n
ts
.
I
t
b
eg
in
s
with
a
c
o
n
v
o
lu
tio
n
al
lay
er
f
o
llo
we
d
b
y
im
p
r
o
v
ed
b
atch
n
o
r
m
aliza
tio
n
an
d
s
to
ch
asti
c
p
o
o
lin
g
.
Su
b
s
eq
u
en
tly
,
a
n
o
th
er
co
n
v
o
l
u
tio
n
al
lay
er
is
em
p
lo
y
e
d
,
s
u
cc
ee
d
ed
b
y
im
p
r
o
v
ed
b
atch
n
o
r
m
aliza
tio
n
an
d
av
er
ag
e
p
o
o
lin
g
.
Fo
llo
win
g
t
h
is
,
an
o
th
er
co
n
v
o
lu
tio
n
al
lay
er
with
im
p
r
o
v
ed
b
atch
n
o
r
m
aliza
tio
n
is
u
tili
ze
d
,
f
o
llo
wed
b
y
s
to
ch
asti
c
p
o
o
lin
g
.
T
h
e
m
o
d
el
f
u
r
t
h
er
em
p
lo
y
s
s
ca
le
d
o
t
p
r
o
d
u
ct
o
p
er
atio
n
s
,
an
FC
lay
er
an
d
a
n
im
p
r
o
v
e
d
ac
tiv
atio
n
f
u
n
ctio
n
f
o
r
en
h
a
n
ce
d
p
er
f
o
r
m
a
n
ce
in
class
if
icatio
n
task
s
.
T
h
u
s
,
th
e
d
etailed
p
r
o
ce
d
u
r
al
ex
p
lan
atio
n
f
o
r
ea
c
h
lay
e
r
in
t
h
e
I
L
eNe
t
-
5
m
o
d
el
is
g
iv
en
b
elo
w
an
d
its
p
icto
r
ial
r
ep
r
esen
tatio
n
is
d
e
p
icted
in
Fig
u
r
e
7
.
a.
C
o
n
v
o
lu
tio
n
al
l
ay
e
r
(
co
n
v
o
l
u
tio
n
)
:
T
h
is
lay
er
a
p
p
lies
co
n
v
o
lu
tio
n
al
f
ilter
s
to
th
e
in
p
u
t
r
et
r
iev
ed
f
ea
tu
r
es
,
wh
ich
allo
ws
th
e
n
etwo
r
k
t
o
lear
n
s
p
atial
h
ier
a
r
ch
ies
o
f
f
ea
tu
r
es.
E
ac
h
f
ilter
co
n
v
o
lv
e
s
ac
r
o
s
s
th
e
in
p
u
t im
ag
e,
ca
p
tu
r
in
g
d
if
f
er
e
n
t p
atter
n
s
.
b.
I
m
p
r
o
v
ed
b
atc
h
n
o
r
m
aliza
tio
n
:
I
t
n
o
r
m
alize
s
th
e
ac
tiv
atio
n
s
[
4
2
]
o
f
th
e
p
r
ev
io
u
s
lay
er
ac
r
o
s
s
th
e
m
in
i
-
b
atch
.
E
q
u
atio
n
(
2
3
)
e
x
p
lain
s
th
e
co
n
v
e
n
tio
n
al
b
atc
h
n
o
r
m
al
izatio
n
.
,
ℎ
,
,
←
ℎ
,
ℎ
,
,
−
ℎ
√
ℎ
2
+
∈
+
ℎ
∀
,
ℎ
,
,
(
2
3
)
Evaluation Warning : The document was created with Spire.PDF for Python.