I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
3
,
J
une
2025
, pp.
2291
~
2301
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
3
.pp
2291
-
2301
2291
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
S
k
i
n
c
an
c
e
r
d
i
agn
os
i
s u
si
n
g h
y
b
r
i
d
d
e
e
p
p
r
e
-
t
r
ai
n
e
d
c
on
vol
u
t
i
on
al
n
e
u
r
al
n
e
t
w
or
k
s
M
ah
a I
b
r
ah
im
K
h
al
e
e
l
1,
2
, A
m
ir
L
ak
iz
ad
e
h
1
1
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
E
ngi
ne
e
r
i
ng a
nd I
nf
o
r
m
a
t
i
on T
e
c
hnol
ogy, U
ni
ve
r
s
i
t
y
of
Q
om
, Q
om
, I
r
a
n
2
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
T
e
c
hni
que
s
E
ngi
ne
e
r
i
ng, A
l
s
a
f
w
a
U
ni
ve
r
s
i
t
y C
ol
l
e
g
e
,
K
e
r
ba
l
a
, I
r
a
q
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
A
pr
4, 2024
R
e
vi
s
e
d
F
e
b 3, 2025
A
c
c
e
pt
e
d
M
a
r
15, 2025
As
a
variant
of
skin
cancer,
melanoma
represents
a
substantial
men
ace
to
the
health
and
overall
well
-
being
of
individuals.
Statistics
reveal
that
5
5%
of
skin
cancer
patients
succumb
to
this
particular
disease.
However,
early
detection
plays
a
crucial
role
in
reducing
mortality
rates
and
saving
li
ves.
In
the
past
several
decades,
there
has
been
a
rise
in
the
adoption
o
f
deep
learning
algorithms,
capturing
the
interest
of
resear
chers
working
in
this
field.
One
popular
method
involves
utilizing
pre
-
trained
dee
p
neural
networks.
In
this
study,
a
hybrid
approach
is
employed
to
extract
features
from
melanoma
images.
This
approach
integrat
es
the
utilization
o
f
pre
-
trained
architec
tures,
including
AlexNet,
ResNet
-
50,
and
GoogleNet. During
the
transfer
training
phase,
these
networks
are
fine
-
tuned
to
detec
t
skin
cancer
by
adjusting
the
learning
rate.
Subsequently,
the
maximum
rel
evance
minimum
redundanc
y
(
MRMR
)
algorit
hm
is
employed
to
select
o
ptimal
features
based
on
the
concepts
of
minimum
redundancy
and
ma
ximum
rel
evance
in
order
to
minimize
feature
redundancy
and
e
nhance
classifi
cation
accuracy.
The
bagging
techniqu
e
is
employed
f
or
the
classifi
cation
of
various
skin
cancer
types.
The
experiment
al
results
demonstrate
the
success
of
the
suggested
approach,
yielding
98.9%
accuracy.
Furtherm
ore,
the
results
indicate
the
superiori
ty
of
this
method
according
to
precision,
reca
ll,
and
F
1
-
score
in
comparison
with
e
xisting
algorit
hms.
K
e
y
w
o
r
d
s
:
C
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
F
e
a
tu
r
e
s
e
le
c
ti
on
M
R
M
R
a
lg
or
it
hm
P
r
e
-
tr
a
in
e
d ne
twor
ks
S
ki
n c
a
nc
e
r
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
M
a
ha
I
br
a
hi
m
K
ha
le
e
l
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
E
ngi
ne
e
r
in
g a
nd I
nf
or
m
a
ti
on T
e
c
hnol
ogy, Unive
r
s
it
y of
Q
om
Q
om
, I
r
a
n
E
m
a
il
:
m
.i
br
a
hi
m
@
s
tu
.qom.a
c
.i
r
1.
I
N
T
R
O
D
U
C
T
I
O
N
S
ki
n
c
a
nc
e
r
,
a
pe
r
va
s
iv
e
a
nd
pot
e
nt
ia
ll
y
le
th
a
l
c
la
s
s
of
ne
opl
a
s
ti
c
di
s
or
de
r
s
,
a
r
is
e
s
f
r
om
th
e
unc
ont
r
ol
le
d gr
ow
th
of
s
ki
n c
e
ll
s
, of
te
n t
r
ig
ge
r
e
d by c
um
ul
a
ti
ve
e
xpos
ur
e
t
o ul
tr
a
vi
ol
e
t
(
U
V
)
r
a
di
a
ti
on. W
hi
le
va
r
io
us
s
ki
n
c
a
nc
e
r
s
e
xi
s
t,
th
e
th
r
e
e
pr
im
a
r
y
ty
pe
s
a
r
e
ba
s
a
l
c
e
ll
c
a
r
c
in
om
a
(
B
C
C
)
,
s
qua
m
ous
c
e
ll
c
a
r
c
in
om
a
(
S
C
C
)
,
a
nd
m
e
la
nom
a
[
1]
–
[
3]
.
E
a
c
h
ty
pe
pos
s
e
s
s
e
s
di
s
ti
nc
t
c
ha
r
a
c
te
r
is
ti
c
s
,
hi
s
to
pa
th
ol
ogi
c
a
l
f
e
a
tu
r
e
s
,
a
nd
pr
ognos
e
s
,
ne
c
e
s
s
it
a
ti
ng
a
nua
nc
e
d
unde
r
s
ta
ndi
ng
f
or
a
c
c
ur
a
te
di
a
gnos
i
s
a
nd
ta
r
ge
te
d
th
e
r
a
p
e
ut
ic
in
te
r
ve
nt
io
ns
.
I
n
a
ddi
ti
on
to
th
e
s
e
m
a
jo
r
ty
pe
s
,
num
e
r
ous
ot
he
r
s
ki
n
le
s
io
n
s
m
e
r
it
a
tt
e
nt
io
n,
r
a
ngi
ng
f
r
om
be
ni
gn
m
ol
e
s
a
nd
c
ys
ts
to
pr
e
c
a
nc
e
r
ous
a
c
ti
ni
c
ke
r
a
to
s
is
.
T
he
di
ve
r
s
e
s
pe
c
tr
um
of
s
ki
n
le
s
io
ns
unde
r
s
c
or
e
s
th
e
c
ha
ll
e
nge
of
a
c
c
ur
a
te
di
a
gno
s
is
,
pr
om
pt
in
g
th
e
e
xpl
or
a
ti
on
of
a
dva
nc
e
d
di
a
gnos
ti
c
te
c
hni
que
s
a
nd
te
c
hnol
ogi
e
s
[
4]
,
[
5]
.
T
im
e
ly
di
a
gnos
is
c
on
s
ti
tu
te
s
a
pi
vot
a
l
r
ol
e
in
th
e
s
uc
c
e
s
s
f
ul
m
a
na
ge
m
e
nt
a
nd
in
te
r
ve
nt
io
n
of
s
ki
n
c
a
nc
e
r
,
unde
r
s
c
or
in
g
c
r
it
ic
a
l
ne
e
d
f
or
a
c
c
ur
a
te
a
nd
e
f
f
ic
ie
nt
di
a
gnos
ti
c
m
e
th
od
s
.
H
is
to
r
ic
a
ll
y,
vi
s
ua
l
in
s
pe
c
ti
on
by
de
r
m
a
to
lo
gi
s
ts
ha
s
be
e
n
th
e
pr
im
a
r
y
m
e
a
ns
of
di
a
gnos
in
g
s
ki
n
le
s
io
ns
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 3, J
une
2025
:
2291
-
2301
2292
H
ow
e
ve
r
,
th
e
in
he
r
e
nt
s
ubj
e
c
ti
vi
ty
a
nd
pot
e
nt
ia
l
f
or
hum
a
n
e
r
r
or
in
th
is
a
ppr
oa
c
h
ha
v
e
pr
om
pt
e
d
th
e
e
xpl
or
a
ti
on
of
a
dva
nc
e
d
te
c
hnol
ogi
e
s
a
nd
m
e
th
odol
ogi
e
s
to
e
nha
nc
e
di
a
gnos
ti
c
pr
e
c
is
io
n
[
6]
.
B
io
ps
ie
s
a
r
e
c
om
m
onl
y
us
e
d
to
d
e
te
c
t
s
ki
n
c
a
nc
e
r
,
in
vol
vi
ng
th
e
r
e
m
ova
l
o
f
a
s
pe
c
im
e
n
f
r
om
a
pot
e
nt
ia
ll
y
a
f
f
li
c
te
d
a
r
e
a
on
th
e
s
ki
n
f
or
m
e
di
c
a
l
te
s
ti
ng
to
c
onf
ir
m
it
s
c
a
nc
e
r
ous
na
tu
r
e
.
T
hi
s
m
e
th
od,
w
hi
le
e
f
f
e
c
ti
ve
,
is
in
va
s
iv
e
,
di
s
c
om
f
or
ti
ng,
a
nd
de
m
a
nds
a
s
ig
ni
f
ic
a
nt
a
m
ount
of
t
im
e
.
T
he
a
dve
nt
of
c
om
put
a
ti
ona
l
te
c
hnol
ogi
e
s
of
f
e
r
s
a
f
a
s
te
r
,
m
or
e
c
onve
ni
e
nt
,
a
nd
c
os
t
-
e
f
f
e
c
ti
ve
a
lt
e
r
na
ti
ve
f
or
di
a
gnos
in
g
s
ki
n
c
a
nc
e
r
s
ym
pt
om
s
.
T
he
s
e
s
y
s
te
m
s
a
im
to
c
onduc
t
a
n
in
it
ia
l
e
va
lu
a
ti
on
of
pot
e
nt
ia
ll
y
pr
obl
e
m
a
ti
c
s
ki
n
le
s
io
ns
ut
il
iz
in
g
hi
gh
-
r
e
s
ol
ut
io
n,
m
e
di
c
a
l
da
ta
a
nd
m
a
c
hi
n
e
le
a
r
ni
ng
(
ML
)
te
c
hni
que
s
.
A
lt
hough
th
e
s
e
s
ys
te
m
s
c
a
nnot
r
e
pl
a
c
e
th
e
c
r
uc
ia
l
f
unc
ti
on
of
hi
s
to
pa
th
ol
ogy
in
c
a
nc
e
r
de
te
c
ti
on,
th
e
de
ve
lo
pm
e
nt
of
s
ophi
s
ti
c
a
te
d
de
te
c
ti
on
te
c
hni
que
s
c
a
n
s
ub
s
ta
nt
ia
ll
y
a
s
s
is
t
in
th
e
e
a
r
ly
de
te
c
ti
on
of
m
a
li
gna
nt
tu
m
or
s
,
pa
r
ti
c
ul
a
r
ly
in
r
e
gi
ons
w
it
h
li
m
it
e
d
a
c
c
e
s
s
to
m
e
di
c
a
l
r
e
s
our
c
e
s
a
nd he
a
lt
hc
a
r
e
pr
of
e
s
s
io
na
l
s
[
7]
–
[
9]
.
N
um
e
r
ous
non
-
in
va
s
iv
e
m
e
th
ods
ha
ve
s
pur
r
e
d
r
e
s
e
a
r
c
he
r
s
to
de
te
c
t
s
ig
ns
of
s
ki
n
c
a
nc
e
r
a
nd
a
s
c
e
r
ta
in
w
he
th
e
r
th
e
y
in
di
c
a
te
th
e
pr
e
s
e
nc
e
of
m
e
la
nom
a
.
G
iv
e
n
it
s
non
-
in
va
s
iv
e
c
ha
r
a
c
te
r
is
ti
c
s
,
im
a
ge
pr
oc
e
s
s
in
g
h
a
s
b
e
e
n
w
id
e
ly
r
e
c
om
m
e
nde
d
f
or
th
e
de
te
c
ti
on
of
m
e
la
nom
a
.
I
t
ha
s
e
vol
v
e
d
a
s
a
va
lu
a
bl
e
de
te
c
ti
on
m
e
th
od
f
or
th
e
a
na
ly
s
is
of
c
li
ni
c
a
l
d
a
ta
,
f
a
c
il
it
a
ti
ng
pr
om
pt
a
nd
e
f
f
ic
a
c
io
us
m
e
di
c
a
l
in
te
r
ve
nt
io
ns
f
or
pa
ti
e
nt
s
.
I
m
a
ge
pr
oc
e
s
s
in
g
e
qui
ps
pr
of
e
s
s
io
n
a
ls
w
it
h
a
r
e
a
l
-
ti
m
e
de
te
c
ti
on
m
e
th
od,
f
a
c
il
it
a
ti
ng
th
e
m
oni
to
r
in
g
of
pa
ti
e
nt
’
s
c
ondi
ti
on
ove
r
ti
m
e
,
s
uppl
yi
ng
im
a
ge
s
f
or
tr
a
in
in
g
a
nd
de
m
ons
tr
a
ti
on,
a
nd
e
na
bl
in
g
s
w
if
t
im
a
ge
c
om
pa
r
is
on.
T
hi
s
c
oul
d
a
ls
o
pr
ove
to
be
c
o
s
t
-
e
f
f
ic
ie
nt
f
or
hos
pi
ta
ls
[
10]
–
[
12]
.
D
e
r
m
os
c
opy,
e
na
bl
in
g non
-
s
ur
gi
c
a
l
s
c
r
ut
in
y of
th
e
hypode
r
m
is
, c
a
n pr
oduc
e
c
om
m
e
nda
bl
e
out
c
om
e
s
;
how
e
ve
r
, i
t
de
m
a
nds
e
xt
e
ns
iv
e
tr
a
in
in
g
a
nd
e
xpe
r
ti
s
e
in
de
r
m
a
to
lo
gy
f
or
e
f
f
e
c
ti
ve
a
ppl
ic
a
ti
on.
R
e
gr
e
tt
a
bl
y,
th
e
m
e
th
od
f
a
il
s
to
f
ur
ni
s
h
a
de
f
in
it
iv
e
m
e
la
nom
a
de
te
c
ti
on,
pa
r
ti
c
ul
a
r
ly
a
t
th
e
in
it
ia
l
pha
s
e
s
.
C
on
s
e
que
nt
ly
,
th
e
ne
e
d
f
or
a
n
a
ut
om
a
te
d
di
a
gnos
ti
c
to
ol
b
e
c
om
e
s
in
di
s
pe
n
s
a
bl
e
.
ML
,
a
br
oa
d
a
nd
a
dva
nc
in
g s
ubs
e
t
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
,
ha
s
be
e
n
e
xt
e
ns
iv
e
ly
in
c
or
por
a
te
d
in
to
c
ont
e
m
por
a
r
y
c
om
put
e
r
-
ba
s
e
d
te
c
hnol
ogi
e
s
.
T
hi
s
te
c
hnol
ogy,
w
hi
c
h a
ll
ow
s
c
om
put
e
r
s
t
o l
e
a
r
n a
nd e
vol
ve
a
ut
onomous
ly
w
it
hout
e
xpl
ic
it
pr
ogr
a
m
m
in
g, o
r
ig
in
a
te
s
f
r
om
t
he
s
tu
dy
of
pa
tt
e
r
n
r
e
c
ogni
ti
on
a
nd
c
om
put
a
ti
ona
l
l
e
a
r
ni
ng
th
e
or
y.
O
ve
r
th
e
p
a
s
t
f
e
w
de
c
a
d
e
s
,
it
ha
s
f
ound
it
s
a
ppl
ic
a
ti
on
in
a
m
yr
ia
d
of
dom
a
in
s
[
13]
,
[
14
]
.
I
n
r
e
s
pons
e
to
th
e
s
e
c
ha
ll
e
ng
e
s
,
th
e
in
te
gr
a
ti
on
of
A
I
a
nd
de
e
p
le
a
r
ni
ng
a
lg
or
it
hm
s
ha
s
s
ur
f
a
c
e
d
a
s
a
c
om
pe
ll
in
g
a
ppr
oa
c
h
to
e
nha
nc
e
th
e
e
f
f
ic
ie
nc
y
a
nd
pr
e
c
is
io
n
of
s
ki
n
c
a
nc
e
r
de
te
c
ti
on.
C
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
s
)
a
r
e
a
s
ub
s
e
t
of
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
ks
th
a
t
a
r
e
e
xt
e
ns
iv
e
ly
us
e
d
f
or
pr
oc
e
s
s
in
g
vi
s
ua
l
da
ta
.
T
he
y
ope
r
a
te
b
y
a
c
qui
r
in
g
th
e
di
s
ti
nc
ti
ve
f
e
a
tu
r
e
s
f
r
om
th
e
tr
a
in
in
g
da
ta
s
e
t
f
or
c
la
s
s
if
yi
ng
th
e
te
s
t
da
ta
s
e
t,
e
m
pl
oyi
ng
a
pr
oc
e
s
s
in
vol
vi
ng
ba
c
kpr
opa
ga
ti
on
a
nd
f
e
e
df
or
w
a
r
d,
a
ki
n
to
ot
he
r
a
r
ti
f
ic
ia
l
ne
u
r
a
l
ne
twor
ks
.
W
hi
le
C
N
N
s
out
pe
r
f
or
m
ba
s
ic
ML
m
e
th
ods
in
te
r
m
s
o
f
pe
r
f
or
m
a
nc
e
,
th
e
y
r
e
qui
r
e
m
or
e
c
om
put
a
ti
ona
l
r
e
s
our
c
e
s
.
T
he
c
om
pl
e
xi
ty
of
th
e
ne
twor
k
a
nd
it
s
a
bi
li
ty
to
id
e
nt
if
y pa
tt
e
r
ns
i
nc
r
e
a
s
e
w
it
h t
he
numbe
r
of
l
a
ye
r
s
[
15]
–
[
17]
.
T
he
pr
e
s
e
nt
in
ve
s
ti
ga
ti
on
in
tr
oduc
e
s
a
nove
l
m
ode
l
in
de
te
c
ti
n
g
s
ki
n
c
a
nc
e
r
by
pr
opos
in
g
a
hybr
id
de
e
p
pr
e
-
tr
a
in
e
d
C
N
N
a
r
c
hi
te
c
tu
r
e
.
T
he
hybr
id
m
ode
l
in
te
gr
a
te
s
th
e
s
tr
e
ngt
hs
of
va
r
io
us
pr
e
-
tr
a
in
e
d
C
N
N
s
,
opt
im
iz
in
g
th
e
le
a
r
ni
ng
pr
oc
e
s
s
f
or
de
r
m
os
c
opi
c
im
a
ge
s
.
T
he
m
a
jo
r
c
ont
r
ib
ut
io
ns
a
r
e
:
i)
a
c
om
bi
na
ti
on
of
th
r
e
e
pr
e
-
tr
a
in
e
d
de
e
p
le
a
r
ni
ng
m
ode
l
s
in
c
lu
di
ng
A
l
e
xN
e
t,
R
e
s
N
e
t5
0
,
a
nd
G
oogl
e
N
e
t
to
e
xt
r
a
c
t
a
n
e
f
f
e
c
ti
ve
s
e
t
of
f
e
a
tu
r
e
s
f
r
om
s
ki
n
im
a
ge
s
;
ii
)
u
s
in
g
th
e
tr
a
n
s
f
e
r
le
a
r
ni
ng
a
ppr
oa
c
h,
in
w
hi
c
h
pr
e
-
tr
a
in
e
d
A
le
xN
e
t,
R
e
s
N
e
t5
0
,
a
nd
G
oogl
e
N
e
t
is
tr
a
in
e
d
to
de
te
c
t
s
ki
n
c
a
nc
e
r
by
r
e
duc
in
g
th
e
ne
twor
k'
s
le
a
r
ni
ng
r
a
te
a
t
th
e
f
in
e
-
tu
ni
ng
s
ta
ge
;
ii
i)
a
c
hi
e
vi
ng
hi
gh
de
te
c
ti
on
a
c
c
ur
a
c
y
in
li
m
it
e
d
num
be
r
of
s
ki
n
c
a
nc
e
r
in
put
im
a
ge
s
a
nd
r
e
duc
in
g
de
te
c
ti
on
e
r
r
or
due
to
th
e
us
e
of
f
in
e
-
tu
ni
ng
m
e
th
od
;
a
nd
iv
)
s
e
le
c
ti
ng
th
e
opt
im
a
l
f
e
a
tu
r
e
s
u
s
in
g
th
e
two
c
onc
e
pt
s
of
m
in
im
um
r
e
dunda
nc
y
a
nd
m
a
xi
m
um
r
e
le
va
nc
e
,
in
or
de
r
to
r
e
duc
e
r
e
dunda
nc
y
be
twe
e
n
f
e
a
tu
r
e
s
a
nd i
nc
r
e
a
s
e
c
la
s
s
if
ic
a
ti
on a
c
c
ur
a
c
y.
T
he
r
e
m
a
in
de
r
of
th
is
pa
pe
r
is
or
ga
ni
z
e
d
in
th
e
f
ol
lo
w
in
g
m
a
n
ne
r
.
S
e
c
ti
on
2
in
tr
oduc
e
s
th
e
r
e
la
te
d
w
or
ks
. S
e
c
ti
on
3
e
la
bor
a
te
s
on t
he
pr
opos
e
d m
e
th
odol
ogy
. S
e
c
t
io
n
4
di
s
c
lo
s
e
s
t
h
e
c
onduc
te
d e
xp
e
r
im
e
nt
s
a
nd
th
e
ir
r
e
s
pe
c
ti
ve
r
e
s
ul
ts
. F
in
a
ll
y,
s
e
c
ti
on
5
c
onc
lu
de
s
t
he
pa
pe
r
.
2.
R
E
L
A
T
E
D
WORKS
K
e
e
r
th
a
na
e
t
a
l
.
[
18]
in
tr
oduc
e
th
e
in
t
e
gr
a
ti
on
of
C
N
N
a
n
d
a
s
uppor
t
ve
c
to
r
m
a
c
hi
n
e
(
S
V
M
)
c
la
s
s
if
i
e
r
,
two
in
nov
a
ti
ve
h
ybr
id
m
od
e
l
s
.
T
he
s
e
m
od
e
l
s
a
r
e
d
e
s
ig
ne
d
to
c
a
t
e
gor
iz
e
de
r
m
os
c
opy
im
a
ge
s
a
s
e
it
he
r
b
e
ni
gn
or
m
e
l
a
nom
a
le
s
io
n
s
.
T
h
e
S
V
M
c
la
s
s
if
ie
r
r
e
c
e
i
v
e
s
c
onc
a
te
na
t
e
d
f
e
a
tu
r
e
s
e
x
tr
a
c
t
e
d
by
th
e
f
ir
s
t
a
nd
s
e
c
o
nd
C
N
N
m
ode
l
s
f
or
th
e
c
l
a
s
s
if
i
c
a
ti
o
n
ta
s
k.
T
he
p
e
r
f
or
m
a
nc
e
of
th
e
s
e
pr
opo
s
e
d
m
e
th
o
ds
i
s
e
v
a
lu
a
t
e
d
a
ga
in
s
t
la
b
e
ls
pr
ov
id
e
d
by
a
pr
of
e
s
s
io
n
a
l
d
e
r
m
a
to
lo
gi
s
t.
W
h
e
n a
s
s
e
s
s
e
d
u
s
in
g
th
e
p
ubl
ic
l
y a
va
il
a
bl
e
I
S
B
I
201
6
da
ta
b
a
s
e
,
th
e
s
e
m
e
th
od
s
de
m
o
ns
tr
a
te
d
s
upe
r
i
or
p
e
r
f
or
m
a
n
c
e
c
om
pa
r
e
d
to
e
x
is
ti
ng
t
e
c
hni
qu
e
s
.
T
hi
s
s
ug
ge
s
ts
th
a
t
th
e
pr
op
os
e
d f
r
a
m
e
w
or
k
c
ou
ld
s
ig
ni
f
i
c
a
nt
l
y e
n
ha
n
c
e
t
he
a
c
c
ur
a
c
y
of
de
r
m
o
s
c
opy i
m
a
g
e
c
l
a
s
s
if
i
c
a
ti
o
n.
R
a
hm
a
n
e
t
al
.
[
19]
s
ta
te
d
a
m
e
th
od
ba
s
e
d
on a
ni
s
ot
r
opi
c
di
f
f
us
io
n
f
il
te
r
in
g
is
a
ppl
ie
d
to
d
e
r
m
os
c
opy
im
a
ge
s
to
m
it
ig
a
te
noi
s
e
.
T
he
n,
th
e
f
a
s
t
-
bounding
box
(
F
B
B
)
te
c
hni
que
is
e
m
pl
oye
d
f
or
th
e
s
e
gm
e
nt
a
ti
on
of
s
ki
n
c
a
n
c
e
r
r
e
gi
ons
.
T
w
o
f
e
a
tu
r
e
e
xt
r
a
c
to
r
s
a
r
e
us
e
d
f
or
th
e
r
e
pr
e
s
e
nt
a
ti
on
of
th
e
im
a
ge
s
.
T
h
e
in
it
ia
l
f
e
a
tu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Sk
in
c
anc
e
r
di
agno
s
is
us
in
g
h
y
br
id
de
e
p p
r
e
-
tr
ai
ne
d c
on
v
ol
ut
io
nal
ne
ur
al
ne
tw
or
k
s
(
M
aha I
br
ahi
m
K
hal
e
e
l
)
2293
e
xt
r
a
c
to
r
ut
il
iz
e
d
is
th
e
hybr
id
f
e
a
tu
r
e
e
xt
r
a
c
to
r
(
H
F
E
)
,
f
ol
lo
w
e
d
by
a
C
N
N
m
ode
le
d
on
V
G
G
19.
T
he
H
F
E
in
te
gr
a
te
s
th
r
e
e
te
c
hni
que
s
-
s
pe
e
d
up
r
obus
t
f
e
a
tu
r
e
(
S
U
R
F
)
,
hi
s
to
gr
a
m
-
or
ie
nt
e
d
gr
a
di
e
nt
(
H
O
G
)
a
nd
lo
c
a
l
bi
na
r
y
pa
tt
e
r
n
(
L
B
P
)
-
to
ge
ne
r
a
te
a
uni
f
ie
d
f
e
a
tu
r
e
ve
c
to
r
.
T
he
C
N
N
m
e
th
odol
ogy
i
s
th
e
n
a
ppl
ie
d
to
pr
ovi
de
s
uppl
e
m
e
nt
a
r
y
f
e
a
tu
r
e
s
.
S
ubs
e
que
nt
ly
,
th
e
s
e
ve
c
to
r
s
a
r
e
a
m
a
lg
a
m
a
te
d
to
f
or
m
ul
a
te
th
e
c
la
s
s
if
ic
a
ti
on
f
r
a
m
e
w
or
k. A
c
c
or
di
ng t
o t
he
s
im
ul
a
ti
on f
in
di
ngs
, t
he
s
ugge
s
te
d
a
ppr
oa
c
h outpe
r
f
or
m
e
d ot
he
r
m
e
th
ods
.
B
oz
k
ur
t
[
20]
pr
opo
s
e
a
p
ot
e
nt
da
ta
e
nr
ic
hm
e
nt
a
n
d
a
pr
e
tr
a
in
e
d
m
od
e
l
f
or
t
he
s
ki
n
le
s
io
ns
c
la
s
s
if
i
c
a
ti
on.
I
n
c
e
p
ti
on
-
R
e
s
N
et
-
v
2,
a
s
a
n
in
te
gr
a
te
d
a
r
c
hi
t
e
c
tu
r
e
,
i
s
in
tr
oduc
e
d
f
or
th
e
c
a
te
g
or
iz
a
ti
on
of
c
a
n
c
e
r
ou
s
im
a
ge
s
.
T
he
pr
im
a
r
y
obj
e
c
ti
v
e
is
a
ugm
e
nt
in
g
th
e
nu
m
be
r
of
im
a
g
e
s
in
t
he
da
ta
s
e
t
l
e
ve
r
a
gi
ng
d
a
ta
a
ugm
e
nt
a
ti
o
n a
nd e
x
a
m
in
in
g
it
s
i
nf
lu
e
n
c
e
on
th
e
m
od
e
l.
W
it
h
t
he
a
ugm
e
nt
e
d
d
a
ta
s
e
t,
th
e
I
n
c
e
pt
io
n
-
R
e
s
N
et
-
v2
m
ode
l
a
c
hi
e
ve
d t
he
hi
ghe
s
t
r
e
c
or
de
d a
c
c
ur
a
c
y.
M
a
q
s
oo
d
a
nd
D
a
m
a
š
e
vi
č
iu
s
[
21]
in
tr
oduc
e
a
n
e
xt
e
n
s
iv
e
c
om
put
e
r
-
a
id
e
d
di
a
gn
os
i
s
(
C
A
D
)
m
o
de
l
f
ounde
d
on a
de
e
p a
r
c
hi
t
e
c
tu
r
e
de
s
ig
ne
d f
or
bot
h t
he
c
la
s
s
if
ic
a
t
i
on a
nd
s
e
gm
e
nt
a
ti
o
n of
s
ki
n l
e
s
io
n
s
. I
ni
ti
a
ll
y,
a
pr
e
-
pr
oc
e
s
s
in
g i
s
p
e
r
f
or
m
e
d
ut
il
iz
i
ng a
n
a
ppr
o
a
c
h
th
a
t
i
nvol
v
e
s
m
odi
f
ie
d bi
o
-
in
s
pi
r
e
d m
ul
ti
pl
e
e
xp
os
ur
e
f
u
s
io
n
f
or
c
ont
r
a
s
t
e
nh
a
nc
e
m
e
nt
.
F
ol
l
ow
in
g
th
i
s
,
a
s
pe
c
ia
li
z
e
d
C
N
N
a
r
c
hi
te
c
tu
r
e
c
om
pr
i
s
in
g
26
la
y
e
r
s
i
s
c
r
e
a
te
d
to
pe
r
f
or
m
th
e
s
e
gm
e
nt
a
ti
on
of
s
ki
n
r
e
gi
on
s
.
I
n
th
e
s
ub
s
e
q
ue
nt
p
ha
s
e
,
f
o
ur
pr
e
-
tr
a
in
e
d
C
N
N
m
od
e
l
s
(
V
G
G
16,
R
e
s
N
e
t
-
50,
X
c
e
p
ti
on
,
a
nd R
e
s
N
e
t
-
101)
a
r
e
a
dj
u
s
te
d
a
nd
tr
a
in
e
d
th
r
ough
tr
a
ns
f
e
r
le
a
r
ni
ng us
i
ng
th
e
s
e
gm
e
nt
e
d
im
a
ge
s
.
T
h
e
f
e
a
tu
r
e
s
e
le
c
ti
on
pr
oc
e
s
s
ut
il
iz
e
s
a
P
oi
s
s
on
di
s
tr
i
b
ut
io
n
a
ppr
oa
c
h
t
o
id
e
nt
if
y
o
pt
im
a
l
f
e
a
tu
r
e
s
f
or
c
la
s
s
if
i
c
a
ti
on.
T
he
ul
ti
m
a
te
c
a
te
gor
iz
a
ti
on
i
s
c
a
r
r
ie
d
out
th
r
oug
h
th
e
a
ppl
i
c
a
ti
o
n
of
th
e
s
e
le
c
te
d
f
e
a
tu
r
e
s
in
th
e
m
ul
ti
-
c
la
s
s
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
M
C
-
S
V
M
)
.
I
n
c
om
p
a
r
is
o
n
to
e
s
t
a
bl
i
s
he
d
s
t
a
te
-
of
-
th
e
-
a
r
t
a
lg
or
it
hm
s
,
th
e
s
ugg
e
s
t
e
d
a
ppr
o
a
c
h f
or
s
k
in
l
e
s
io
n de
te
c
ti
on
a
nd
c
la
s
s
if
ic
a
ti
on
e
xhi
bi
te
d s
upe
r
io
r
pe
r
f
or
m
a
nc
e
.
A
ki
la
nda
s
ow
m
ya
e
t
al
.
[
22]
ut
il
iz
e
a
te
c
hni
que
known
a
s
s
a
nd
c
a
t
s
w
a
r
m
opt
im
iz
a
ti
on
w
it
h
R
e
s
N
e
t5
0 (
S
C
S
O
-
R
e
s
N
e
t5
0)
t
o di
f
f
e
r
e
nt
ia
te
be
twe
e
n de
e
p hi
dde
n a
nd known f
e
a
tu
r
e
s
, t
he
r
e
by e
nha
nc
in
g t
he
a
c
c
ur
a
c
y
of
pr
e
di
c
ti
ons
.
A
ut
hor
s
e
m
pl
oy
a
n
e
nha
nc
e
d
h
a
r
m
ony
s
e
a
r
c
h
(
E
H
S
)
s
tr
a
te
gy
f
or
f
e
a
tu
r
e
opt
im
iz
a
ti
on a
nd r
e
duc
ti
on of
da
ta
di
m
e
ns
io
na
li
ty
. A
r
a
nge
of
e
ns
e
m
bl
e
c
la
s
s
if
ie
r
s
, i
nc
lu
di
ng S
V
M
, k
-
ne
a
r
e
s
t
ne
ig
hbor
(
K
N
N
)
,
r
a
ndom
f
or
e
s
t,
li
ne
a
r
r
e
gr
e
s
s
io
n
,
a
nd
na
iv
e
B
a
ye
s
a
r
e
us
e
d
f
or
th
e
ti
m
e
ly
de
te
c
ti
on
of
s
ki
n
c
a
nc
e
r
.
T
he
e
f
f
e
c
ti
ve
ne
s
s
of
th
e
s
ugg
e
s
te
d
a
ppr
oa
c
h
is
e
va
lu
a
te
d
ba
s
e
d
on
th
e
K
a
ggl
e
a
nd
th
e
I
S
I
C
2019
da
ta
ba
s
e
s
.
A
s
pe
r
e
va
lu
a
ti
ons
on
be
nc
hm
a
r
k
da
ta
s
e
t
s
,
th
e
s
ugge
s
te
d
m
e
th
od
e
nha
nc
e
s
a
c
c
ur
a
c
y
in
c
om
pa
r
is
on
to
le
a
di
ng
te
c
hni
que
s
.
T
he
s
e
r
e
s
ul
t
s
il
lu
s
tr
a
te
th
e
c
a
pa
bi
li
ty
of
th
e
pr
opos
e
d
m
ode
l
in
im
pr
ovi
ng
th
e
t
im
e
ly
de
te
c
ti
on of
s
ki
n c
a
nc
e
r
.
C
ha
nda
e
t
al
.
[
23]
in
tr
oduc
e
a
uni
que
e
ns
e
m
bl
e
a
ppr
oa
c
h
th
a
t
i
nc
or
por
a
te
s
th
r
e
e
de
e
p
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
D
C
N
N
)
,
e
a
c
h
c
us
to
m
iz
e
d
w
it
h
va
r
yi
ng
dr
opout
la
ye
r
s
to
e
nha
nc
e
le
a
r
ni
ng.
C
ons
e
que
nt
ly
,
th
e
s
ugge
s
te
d
f
r
a
m
e
w
or
k,
D
C
E
N
S
ne
t,
a
tt
a
in
s
a
n
e
xc
e
pt
io
na
l
bi
a
s
-
va
r
ia
nc
e
ba
l
a
nc
e
.
U
pon
a
s
s
e
s
s
m
e
nt
u
s
in
g
th
e
w
id
e
ly
-
ut
il
iz
e
d
H
A
M
10000
s
ki
n
le
s
io
n
da
ta
s
e
t,
th
e
pr
opos
e
d
m
ode
l
s
ur
pa
s
s
e
s
e
xi
s
ti
ng
m
e
th
ods
,
s
how
in
g
pot
e
nt
ia
l
to
e
nha
nc
e
t
he
a
c
c
ur
a
c
y of
di
a
gnos
i
s
a
nd t
r
e
a
tm
e
nt
.
A
pi
one
e
r
in
g
m
e
th
od
is
s
ugge
s
t
e
d
f
or
m
ul
ti
-
c
la
s
s
c
la
s
s
if
ic
a
ti
on
of
s
ki
n
le
s
io
ns
,
c
a
pi
ta
li
z
in
g
on
opt
im
a
l
de
e
p l
e
a
r
ni
ng f
e
a
tu
r
e
i
nt
e
gr
a
ti
on a
nd a
n
e
xt
r
e
m
e
l
e
a
r
ni
ng ma
c
hi
ne
(
E
L
M
)
[
24]
.
T
he
s
ugge
s
te
d m
ode
l
e
nc
om
pa
s
s
e
s
th
e
f
ol
lo
w
in
g
m
a
in
s
ta
ge
s
:
c
a
pt
ur
in
g
a
nd
im
pr
ov
in
g
im
a
ge
s
;
e
xt
r
a
c
ti
ng
de
e
p
le
a
r
ni
ng
f
e
a
tu
r
e
s
th
r
ough
tr
a
ns
f
e
r
le
a
r
ni
ng;
c
hoos
in
g
th
e
m
os
t
e
f
f
e
c
ti
ve
f
e
a
tu
r
e
s
ut
il
iz
in
g
a
c
om
bi
ne
d
a
ppr
oa
c
h
of
w
ha
le
opt
im
iz
a
ti
on
a
nd
e
nt
r
opy
-
m
ut
ua
l
in
f
or
m
a
ti
on
(
E
M
I
)
;
a
nd
m
e
r
gi
ng
s
e
le
c
te
d
f
e
a
tu
r
e
s
us
in
g
a
n
a
da
pt
e
d
c
a
noni
c
a
l
c
or
r
e
la
ti
on
-
ba
s
e
d
m
e
th
od
a
nd,
ul
ti
m
a
te
ly
,
c
la
s
s
if
ic
a
t
io
n
a
c
c
or
di
ng
to
E
L
M
.
T
h
e
s
ta
g
e
of
s
e
le
c
ti
ng
f
e
a
tu
r
e
s
e
nha
nc
e
s
th
e
pr
e
c
is
io
n
a
nd
e
f
f
ic
a
c
y
of
th
e
s
ys
te
m
.
I
n
c
ont
r
a
s
t
to
nove
l
te
c
hni
que
s
,
th
e
s
ugge
s
te
d
a
ppr
oa
c
h e
nha
nc
e
s
pr
e
c
is
io
n w
hi
le
m
a
in
ta
in
in
g
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y.
B
a
ns
a
l
e
t
al
.
[
25]
s
ugge
s
t
e
m
pl
oyi
ng
in
te
gr
a
te
d
e
xt
r
a
c
te
d
f
e
a
tu
r
e
s
,
ha
r
ne
s
s
in
g
bot
h
de
e
p
le
a
r
ni
ng
m
ode
l
a
nd
ha
ndc
r
a
f
te
d
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
t
e
c
hni
que
s
,
d
e
r
iv
e
d
f
r
om
de
r
m
os
c
opi
c
im
a
ge
s
to
a
ugm
e
nt
th
e
c
la
s
s
if
ie
r
'
s
f
unc
ti
ona
li
ty
.
E
f
f
ic
ie
nt
N
e
t
-
B
0
a
nd
R
e
s
N
e
t5
0V
2
a
r
e
th
e
n
ut
il
iz
e
d
f
or
e
xt
r
a
c
ti
ng
f
e
a
tu
r
e
s
,
w
hi
le
a
n
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
k
(
ANN
)
is
e
m
pl
oye
d
f
or
c
la
s
s
if
yi
ng
s
ta
ge
.
T
he
out
c
om
e
s
r
e
ve
a
l
th
a
t
th
e
pr
e
s
e
nt
e
d
ha
ir
r
e
m
ova
l
te
c
hni
que
s
,
w
he
n
c
om
bi
n
e
d
w
it
h
th
e
f
e
a
tu
r
e
s
,
e
nha
nc
e
m
e
la
nom
a
di
a
gno
s
is
a
c
c
ur
a
c
y
in
c
om
pa
r
is
on
w
it
h
s
c
e
na
r
io
s
w
h
e
r
e
i
)
no
pr
e
-
pr
oc
e
s
s
in
g
i
s
pe
r
f
or
m
e
d
or
ii
)
ha
ndc
r
a
f
te
d
or
de
e
p
le
a
r
ni
ng
f
e
a
tu
r
e
s
a
r
e
ut
il
iz
e
d i
nde
pe
nde
nt
ly
.
T
he
s
ugg
e
s
te
d m
ode
l
out
pe
r
f
or
m
s
ot
he
r
t
e
c
hni
que
s
i
n t
hi
s
f
ie
ld
.
E
lG
ha
ny
e
t
al
.
[
2
6]
e
xpl
or
e
th
a
t
th
e
po
te
nt
i
a
l
of
de
e
p
s
tr
u
c
tu
r
e
s
i
n
d
e
te
c
ti
ng
va
r
io
us
de
r
m
o
s
c
opi
c
im
a
ge
s
.
T
he
obj
e
c
ti
v
e
i
s
t
o d
e
s
ig
n a
nd a
dj
us
t
a
de
e
p l
e
a
r
ni
ng
a
r
c
hi
te
c
t
ur
e
f
or
di
a
gno
s
in
g v
a
r
io
u
s
gr
a
d
e
s
of
s
ki
n
c
a
n
c
e
r
.
F
i
ne
-
tu
n
in
g,
a
p
ot
e
nt
te
c
hni
q
ue
f
or
a
c
hi
e
v
in
g
im
pr
ove
d
out
c
om
e
s
,
i
s
a
p
pl
ie
d
to
pr
e
tr
a
i
ne
d
m
od
e
l
.
T
h
e
f
in
e
-
tu
ni
ng
of
th
e
pr
opo
s
e
d
d
e
e
p
n
e
twor
k
in
vol
v
e
s
r
e
g
ul
a
r
i
z
a
t
io
n,
ba
t
c
h
nor
m
a
li
z
a
t
io
n,
a
nd
hype
r
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on. T
h
e
s
u
gge
s
te
d
R
e
s
N
e
t5
0
n
e
twor
k e
f
f
e
c
t
iv
e
ly
c
la
s
s
if
ie
d
7
di
s
ti
nc
t
c
l
a
s
s
e
s
of
d
e
r
m
os
c
opi
c
im
a
g
e
s
w
it
hi
n H
A
M
10000
da
t
a
ba
s
e
.
T
he
e
xp
e
r
im
e
nt
a
l
f
i
ndi
ng
s
i
n
di
c
a
te
t
he
s
up
e
r
io
r
it
y of
t
he
s
ug
ge
s
te
d
f
r
a
m
e
w
or
k.
3.
M
E
T
H
O
D
T
he
a
im
of
th
is
s
tu
dy
is
to
pr
opos
e
a
r
obu
s
t
a
lg
or
it
hm
in
or
de
r
to
di
a
gnos
in
g
s
ki
n
c
a
nc
e
r
,
le
ve
r
a
gi
ng
th
e
pow
e
r
of
pr
e
-
tr
a
in
e
d
d
e
e
p
n
e
ur
a
l
ne
twor
ks
.
T
h
e
de
pt
h
a
n
d
e
f
f
e
c
ti
ve
ne
s
s
of
c
onvolut
io
na
l
f
il
te
r
s
m
a
ke
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 3, J
une
2025
:
2291
-
2301
2294
th
e
s
e
pr
e
-
tr
a
in
e
d
C
N
N
s
pa
r
ti
c
ul
a
r
ly
be
ne
f
ic
ia
l
f
or
ta
s
ks
r
e
l
a
te
d
to
im
a
ge
s
.
F
or
th
e
pur
pos
e
of
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
in
th
is
s
tu
dy,
de
e
p
a
r
c
hi
te
c
tu
r
e
s
s
uc
h
a
s
R
e
s
N
e
t,
G
oogl
e
N
e
t
,
a
nd
A
le
xN
e
t
a
r
e
ut
il
iz
e
d.
I
n
th
e
pr
e
s
e
nt
e
d
m
e
th
odol
ogy,
pos
t
th
e
e
xt
r
a
c
ti
on
a
nd
a
m
a
lg
a
m
a
ti
on
of
f
e
a
tu
r
e
s
,
th
e
m
in
im
um
r
e
dunda
nc
y
m
a
xi
m
um
r
e
le
va
nc
e
(
M
R
M
R
)
a
lg
or
it
hm
is
e
m
pl
oye
d
to
pi
c
k
th
e
m
os
t
s
ui
ta
bl
e
f
e
a
tu
r
e
s
.
T
he
M
R
M
R
a
lg
or
it
hm
,
ut
il
iz
in
g
th
e
p
r
in
c
ip
le
s
of
m
in
im
um
r
e
dunda
nc
y
a
nd
m
a
xi
m
um
a
s
s
oc
ia
ti
on,
c
ur
ta
il
s
th
e
r
e
dunda
nc
y
a
m
ongs
t
f
e
a
tu
r
e
s
,
th
e
r
e
by
e
nha
nc
in
g
th
e
a
c
c
ur
a
c
y
of
c
la
s
s
if
ic
a
ti
on.
F
ur
th
e
r
m
or
e
,
th
e
b
a
ggi
ng
a
lg
or
it
hm
is
im
pl
e
m
e
nt
e
d
dur
in
g
th
e
c
la
s
s
if
ic
a
ti
on
pha
s
e
.
F
ig
ur
e
1
il
lu
s
tr
a
te
s
a
bl
oc
k
di
a
gr
a
m
de
pi
c
ti
ng
th
e
pr
opos
e
d m
e
th
odol
ogy
.
F
ig
ur
e
1. D
ia
gr
a
m
of
pr
opos
e
d a
ppr
oa
c
h
3.1.
F
e
at
u
r
e
e
xt
r
ac
t
io
n
I
n
th
e
pr
e
s
e
nt
e
d
a
ppr
oa
c
h,
pr
e
-
t
r
a
in
e
d
ne
twor
ks
s
uc
h
a
s
A
le
xN
e
t,
R
e
s
N
e
t5
0,
a
nd
G
oogl
e
N
e
t
a
r
e
ut
il
iz
e
d f
or
t
he
pu
r
pos
e
of
f
e
a
tu
r
e
e
xt
r
a
c
ti
on f
r
om
i
m
a
ge
s
. D
e
ta
il
e
d de
s
c
r
ip
ti
ons
of
e
a
c
h of
t
he
s
e
ne
twor
ks
a
r
e
pr
ovi
de
d i
n t
he
f
ol
lo
w
in
g s
e
c
ti
ons
.
3.1.1.
A
le
xN
e
t
E
s
s
e
nt
ia
ll
y,
A
le
xN
e
t
is
a
n
a
dva
nc
e
d
ty
pe
of
th
e
tr
a
di
ti
ona
l
L
e
N
e
t.
T
he
c
om
pr
e
he
ns
iv
e
a
r
c
hi
te
c
tu
r
e
of
th
e
A
le
xN
e
t
m
ode
l
is
de
pi
c
te
d
in
ha
d
be
e
n
il
lu
s
tr
a
te
d
F
ig
ur
e
2.
T
he
m
ode
l
c
ons
is
ts
of
s
e
ve
r
a
l
la
ye
r
s
,
s
ta
r
ti
ng
w
it
h
th
e
in
put
la
ye
r
th
a
t
de
te
r
m
in
e
s
th
e
in
put
da
ta
s
iz
e
.
I
n
th
e
A
le
xN
e
t
ne
twor
k
a
r
c
hi
te
c
tu
r
e
,
th
e
s
iz
e
of
th
e
in
put
im
a
ge
s
houl
d
be
227
×
227
×
3.
T
he
he
a
r
t
of
th
e
A
l
e
xN
e
t
is
c
om
pos
e
d
of
th
e
m
id
dl
e
la
y
e
r
s
.
T
he
s
tr
uc
tu
r
e
of
t
he
s
e
l
a
ye
r
s
c
ons
i
s
ts
of
f
iv
e
c
onvolut
io
n l
a
ye
r
s
,
r
e
c
ti
f
ie
d l
in
e
a
r
uni
ts
(
R
e
L
U
)
, m
a
x
-
pool
in
g
la
ye
r
s
,
a
nd f
ul
ly
c
onne
c
te
d l
a
ye
r
s
.
T
he
f
in
a
l
la
ye
r
i
n t
hi
s
s
tr
uc
tu
r
e
i
s
t
h
e
c
la
s
s
if
ic
a
ti
on l
a
y
e
r
.
F
ig
ur
e
2. T
he
a
r
c
hi
te
c
tu
r
e
of
A
le
xN
e
t
m
ode
l
[
27]
3.1.2.
R
e
s
N
e
t
50
R
e
s
id
ua
l
n
e
twor
ks
,
r
e
f
e
r
r
e
d
to
a
s
R
e
s
N
e
ts
,
a
r
e
a
ty
pe
of
de
e
p
c
onvolut
io
na
l
ne
twor
ks
.
T
he
ir
f
unda
m
e
nt
a
l
c
onc
e
pt
in
vol
ve
s
bypa
s
s
in
g
bl
oc
ks
of
c
onvolut
io
na
l
la
ye
r
s
le
ve
r
a
gi
ng
s
hor
tc
ut
c
onne
c
ti
ons
.
T
h
e
ba
s
ic
bl
oc
ks
,
t
e
r
m
e
d
a
s
“
bot
tl
e
n
e
c
k”
bl
oc
k
s
,
a
tt
e
nd
two
f
und
a
m
e
nt
a
l
la
w
e
s
:
i
)
la
ye
r
s
w
it
h
id
e
nt
ic
a
l
out
put
f
e
a
tu
r
e
m
a
p
s
iz
e
s
ha
ve
th
e
s
a
m
e
a
m
ount
of
f
il
te
r
s
a
nd
ii
)
w
he
n
th
e
f
e
a
tu
r
e
m
a
p
di
m
e
ns
io
n
is
c
ut
in
ha
lf
,
th
e
f
il
te
r
c
ount
is
double
d
c
or
r
e
s
pondingl
y.
T
he
dow
n
s
a
m
pl
in
g
pr
oc
e
s
s
is
e
x
e
c
ut
e
d
di
r
e
c
tl
y
by
c
onvolut
io
na
l
la
ye
r
s
w
it
h
a
s
tr
id
e
of
two.
B
a
tc
h
nor
m
a
li
z
a
ti
on
is
im
pl
e
m
e
nt
e
d
a
f
te
r
a
ny
c
onvolut
io
na
l
la
ye
r
a
nd
be
f
or
e
th
e
R
e
L
U
a
c
ti
va
ti
on
f
unc
ti
on.
W
h
e
r
e
th
e
s
iz
e
of
th
e
out
put
a
n
d
in
put
m
a
tc
h,
th
e
id
e
nt
it
y
s
hor
tc
ut
is
us
e
d.
H
ow
e
ve
r
, w
he
r
e
in
t
he
s
iz
e
s
i
nc
r
e
a
s
e
, t
he
pr
oj
e
c
ti
on s
hor
tc
ut
i
s
us
e
d t
o m
a
tc
h s
iz
e
s
vi
a
1*1 c
onvolut
io
ns
. t
he
n
th
e
s
hor
tc
ut
s
go
a
c
r
os
s
f
e
a
tu
r
e
m
a
ps
of
two
di
f
f
e
r
e
nt
s
iz
e
s
.
F
in
a
ll
y,
ne
twor
k
e
nds
by
a
f
ul
ly
c
onne
c
te
d
la
ye
r
w
it
h
1,000 unit
s
, e
qui
ppe
d w
it
h a
S
of
tM
a
x
a
c
ti
va
ti
on f
unc
ti
on.
T
he
ne
twor
k c
ons
i
s
ts
of
a
t
ot
a
l
of
50 we
ig
ht
e
d
la
ye
r
s
, w
it
h 23,534,592 tr
a
in
a
bl
e
va
r
ia
bl
e
s
.
T
he
s
tr
uc
tu
r
e
of
t
he
R
e
s
N
e
t5
0 i
s
i
ll
us
tr
a
te
d i
n F
ig
ur
e
3.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Sk
in
c
anc
e
r
di
agno
s
is
us
in
g
h
y
br
id
de
e
p p
r
e
-
tr
ai
ne
d c
on
v
ol
ut
io
nal
ne
ur
al
ne
tw
or
k
s
(
M
aha I
br
ahi
m
K
hal
e
e
l
)
2295
F
ig
ur
e
3. R
e
s
N
e
t5
0
a
r
c
hi
te
c
tu
r
e
[
27]
3.1.3.
G
oogl
e
N
e
t
A
s
a
w
e
ll
-
e
s
ta
bl
is
he
d
pr
e
tr
a
in
e
d
de
e
p
C
N
N
c
ons
i
s
ti
ng
of
22
la
ye
r
s
,
G
oog
l
e
N
e
t
w
a
s
s
pe
c
if
ic
a
ll
y
de
ve
lo
pe
d
to
out
pe
r
f
or
m
c
ont
e
m
por
a
r
y
C
N
N
ne
twor
ks
in
te
r
m
s
of
pe
r
f
or
m
a
nc
e
.
I
t
de
m
ons
tr
a
te
d
it
s
s
upe
r
io
r
it
y
in
a
c
c
ur
a
te
ly
r
e
c
ogni
z
in
g
vi
s
ua
l
pa
tt
e
r
ns
f
r
om
th
e
s
our
c
e
,
e
a
r
ni
ng
it
th
e
vi
c
to
r
y
in
th
e
I
m
a
ge
N
e
t
la
r
ge
s
c
a
le
vi
s
ua
l
r
e
c
ogni
ti
on
c
ha
ll
e
nge
(
I
L
S
V
R
C
)
c
ha
ll
e
nge
in
2014.
T
he
not
a
bl
e
im
pr
ove
m
e
nt
in
pe
r
f
or
m
a
nc
e
c
a
n
be
c
r
e
di
te
d
to
th
e
in
c
e
pt
io
n
m
odul
e
s
(
I
M
)
in
te
gr
a
te
d
w
it
hi
n
it
s
f
r
a
m
e
w
or
k,
a
s
de
not
e
d
by
th
e
r
e
d c
ir
c
le
in
t
he
i
ll
us
tr
a
ti
on of
F
ig
ur
e
4
.
F
ig
ur
e
4
.
G
oog
l
e
N
e
t
s
tr
uc
tu
r
e
[
27
]
T
he
IM
e
na
bl
e
s
th
e
s
im
ul
ta
ne
ou
s
a
ppl
ic
a
ti
on
of
va
r
io
us
c
onvolut
io
ns
a
nd
m
a
x
pool
in
g
ope
r
a
ti
ons
w
it
hi
n
a
uni
f
ie
d
la
ye
r
.
T
hi
s
gu
a
r
a
nt
e
e
s
th
a
t
th
e
ne
twor
k
un
de
r
goe
s
tr
a
in
in
g
w
it
h
opt
im
iz
e
d
w
e
ig
ht
s
a
nd
c
hoos
e
f
e
a
tu
r
e
s
th
a
t
a
r
e
m
or
e
a
dv
a
nt
a
ge
ou
s
.
T
o
a
c
hi
e
ve
th
is
,
e
a
c
h
in
c
e
pt
io
n
l
a
ye
r
hous
e
s
va
r
ia
bl
e
-
s
iz
e
d
c
onvolut
io
na
l
ke
r
ne
ls
,
s
pe
c
if
ic
a
ll
y
1
×
1,
3
×
3,
a
nd
5
×
5,
a
lo
n
g
w
it
h
a
n
a
ddi
ti
ona
l
3
×
3
m
a
x
pool
in
g.
T
hi
s
a
r
r
a
nge
m
e
nt
is
de
vi
s
e
d
to
e
xt
r
a
c
t
m
or
e
di
s
ti
nc
ti
ve
f
e
a
tu
r
e
s
f
r
o
m
th
e
pa
tt
e
r
n
tr
a
ns
m
it
te
d
f
r
om
th
e
pr
e
c
e
di
ng
la
ye
r
.
T
he
s
tr
uc
tu
r
e
of
th
e
IM
is
de
pi
c
te
d
in
il
lu
s
tr
a
te
d
F
ig
ur
e
3.
H
e
r
e
,
th
e
1*1
f
il
te
r
s
,
in
c
onj
unc
ti
on
w
it
h
th
e
m
a
x
pool
in
g
la
ye
r
,
pe
r
f
or
m
a
dua
l
r
ol
e
of
r
e
duc
in
g
e
nc
a
ps
ul
a
ti
ng
a
nd
di
m
e
ns
io
na
li
ty
th
e
c
ont
e
nt
f
r
om
th
e
pr
e
c
e
di
ng
la
ye
r
.
A
s
il
lu
s
tr
a
te
d
in
F
ig
ur
e
3,
th
e
ni
n
e
I
M
s
in
c
or
por
a
te
d
in
th
e
G
oog
l
e
N
e
t
a
r
c
hi
te
c
tu
r
e
a
r
e
e
f
f
e
c
ti
ve
in
e
xt
r
a
c
ti
ng
m
or
e
di
s
c
r
im
in
a
ti
ve
f
e
a
tu
r
e
s
f
r
om
th
e
s
our
c
e
im
a
ge
.
T
ypi
c
a
ll
y,
f
e
a
tu
r
e
s
w
it
h
gr
e
a
te
r
c
om
pl
e
xi
ty
in
th
e
ir
r
e
pr
e
s
e
nt
a
ti
on
s
a
r
e
f
a
vor
e
d,
gi
ve
n
th
e
ir
e
a
s
e
of
ha
ndl
in
g
in
a
ne
w
ne
twor
k
a
nd s
ubs
ta
nt
ia
l
c
ont
r
ib
ut
io
n t
o e
xpe
di
ti
ng t
he
t
r
a
in
in
g pr
oc
e
s
s
.
3.2.
F
e
at
u
r
e
s
e
le
c
t
io
n
T
he
r
a
ti
ona
le
b
e
hi
nd
e
m
pl
oyi
ng
th
e
M
R
M
R
m
e
th
od
i
s
to
opt
i
m
iz
e
th
e
r
e
le
va
nc
e
be
twe
e
n
f
e
a
tu
r
e
s
a
nd
th
e
ir
c
or
r
e
s
ponding
la
be
ls
(
c
la
s
s
)
,
w
hi
le
c
onc
ur
r
e
nt
ly
a
im
in
g
to
r
e
duc
e
th
e
r
e
dunda
nc
y
a
m
ong
in
di
vi
dua
l
f
e
a
tu
r
e
s
.
T
h
e
M
R
M
R
te
c
hni
qu
e
ut
il
iz
e
s
m
ut
ua
l
in
f
or
m
a
ti
on
a
s
a
m
e
tr
ic
to
a
s
s
e
s
s
th
e
a
s
s
oc
ia
ti
on
a
m
ong
va
r
ia
bl
e
s
. T
he
m
ut
ua
l
in
f
or
m
a
ti
on be
twe
e
n X
a
nd Y
i
s
c
om
put
e
d ba
s
e
d on (
1)
.
(
;
)
=
∑
∑
(
,
)
∈
∈
(
,
)
(
)
(
)
(
1)
(
,
)
de
not
e
s
pr
oba
bi
li
s
ti
c
de
n
s
it
y
of
X
a
nd
Y
va
r
ia
bl
e
s
,
is
in
di
vi
dua
l
c
ha
r
a
c
te
r
is
ti
c
s
,
c
r
e
pr
e
s
e
nt
s
th
e
c
la
s
s
e
s
.
M
a
xi
m
um
r
e
le
va
nc
e
m
e
a
s
ur
e
i
s
c
a
lc
ul
a
te
d ba
s
e
d on (
2)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 3, J
une
2025
:
2291
-
2301
2296
(
,
)
,
=
1
|
|
∑
(
;
)
∈
(
2)
|
S
|
r
e
p
r
e
s
e
n
t
s
t
h
e
di
m
e
n
s
i
on
a
l
i
ty
of
t
h
e
f
e
a
t
ur
e
s
w
it
h
i
n
t
h
e
S
s
p
a
c
e
.
(
;
)
,
m
u
tu
a
l
in
f
or
m
a
ti
o
n
a
m
o
n
g
f
e
a
t
ur
e
s
.
T
he
s
ub
s
e
t
S
,
w
hi
c
h
pos
s
e
s
s
e
s
th
e
ut
m
os
t
c
or
r
e
la
ti
on
r
a
te
,
is
d
e
r
iv
e
d
us
in
g
th
e
m
a
xi
m
um
r
e
le
va
nc
e
m
e
a
s
ur
e
.
A
lt
hough,
it
’
s
c
r
uc
ia
l
to
not
e
th
a
t
th
e
pr
ope
r
ti
e
s
c
hos
e
n
ba
s
e
d
on
th
e
s
e
m
e
a
s
ur
e
s
m
a
y
e
xhi
bi
t
r
e
dunda
nc
y.
H
e
nc
e
,
th
e
ut
il
iz
a
ti
on
of
th
e
m
in
im
um
r
e
dunda
nc
y
c
ondi
ti
on,
a
s
il
lu
s
tr
a
te
d
in
(
3)
,
c
a
n
a
ls
o
s
e
r
ve
a
s
a
m
e
th
od t
o c
hoo
s
e
f
e
a
tu
r
e
s
t
ha
t
e
xhi
bi
t
m
ut
ua
l
e
xc
lu
s
iv
it
y.
(
)
,
=
1
|
|
2
∑
(
,
)
,
∈
(
3)
(
,
)
r
e
pr
e
s
e
nt
s
t
he
m
ut
ua
l
knowle
dge
a
m
ong f
e
a
tu
r
e
s
a
nd
.
T
he
a
im
is
to
e
li
m
in
a
te
hi
ghl
y
de
pe
nde
nt
f
e
a
tu
r
e
s
a
nd
m
in
im
iz
e
r
e
dunda
nc
y
a
m
ong
th
e
m
.
T
he
M
R
M
R
a
lg
or
it
hm
c
om
bi
ne
s
m
a
x
-
r
e
le
va
nc
e
a
nd
m
in
-
r
e
dunda
nc
y
te
r
m
s
.
A
s
a
r
e
s
ul
t,
(
4)
a
nd
(
5)
a
r
e
ut
il
iz
e
d
to
c
onc
ur
r
e
nt
ly
opt
im
iz
e
R
(
r
e
dunda
nc
y)
a
nd D
(
r
e
le
va
nc
e
)
.
(
,
)
,
=
−
(
4)
(
,
)
,
=
(
5)
I
nc
r
e
m
e
nt
a
l
s
e
a
r
c
h
te
c
hni
que
s
a
r
e
ut
il
iz
e
d
f
or
th
e
id
e
nt
if
ic
a
ti
on
of
f
e
a
tu
r
e
s
th
a
t
c
lo
s
e
ly
a
ppr
oxi
m
a
te
opt
im
a
li
ty
in
va
r
io
us
a
ppl
ic
a
ti
on
s
.
T
he
c
a
lc
ul
a
ti
on
of
th
e
opt
i
m
a
l
f
e
a
tu
r
e
s
ub
s
e
t
w
it
h
m
-
1
f
e
a
tu
r
e
s
in
vol
ve
s
th
e
c
om
put
a
ti
on of
t
he
−
1
f
e
a
tu
r
e
s
ubs
e
t,
a
s
out
li
ne
d by (
6)
.
−
−
1
[
(
;
)
−
1
−
1
∑
(
;
)
∈
−
1
(
6)
3.3.
C
la
s
s
if
i
c
at
io
n
I
n
th
is
in
ve
s
ti
ga
ti
on,
th
e
ba
ggi
ng
te
c
hni
que
i
s
e
m
pl
oye
d
f
or
s
k
in
c
a
nc
e
r
de
te
c
ti
on,
w
he
r
e
in
a
s
ubs
e
t
of
th
e
or
ig
in
a
l
da
ta
is
s
e
nt
to
th
e
c
la
s
s
if
ie
r
s
.
C
ons
e
que
nt
ly
,
e
a
c
h
c
la
s
s
if
ie
r
is
e
xpos
e
d
to
a
di
s
ti
nc
t
por
ti
on
o
f
th
e
da
ta
s
e
t,
c
ons
tr
uc
ti
ng
it
s
m
ode
l
ba
s
e
d
on
th
e
s
pe
c
if
ic
s
ub
s
e
t
it
r
e
c
e
iv
e
s
.
N
ot
a
bl
y,
th
e
s
e
le
c
ti
on
of
th
is
s
ubs
e
t
is
pe
r
f
or
m
e
d
w
it
h
r
e
pl
a
c
e
m
e
nt
,
a
ll
ow
in
g
f
or
th
e
pos
s
ib
il
it
y
of
m
ul
t
ip
le
oc
c
ur
r
e
nc
e
s
of
e
a
c
h
s
a
m
pl
e
.
T
he
s
tu
dy
ha
s
s
how
n
th
a
t
th
e
c
la
s
s
if
ic
a
ti
on
m
ode
l
c
a
n
im
pr
ov
e
le
a
r
ni
ng
a
nd
de
te
c
ti
on
c
a
pa
bi
li
ti
e
s
,
r
e
s
ul
ti
ng
in
e
nha
nc
e
d a
c
c
ur
a
c
y. T
h
e
ove
r
a
ll
ope
r
a
ti
on of
t
hi
s
t
e
c
hni
que
i
s
di
s
pl
a
ye
d i
n F
ig
ur
e
5
.
F
ig
ur
e
5. C
la
s
s
if
ic
a
ti
on mode
l
us
in
g ba
ggi
ng a
lg
or
it
hm
4.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
I
n
th
is
pa
r
t
,
th
e
e
f
f
ic
a
c
y
of
th
e
p
r
opos
e
d
m
ode
l
is
e
va
lu
a
te
d
le
v
e
r
a
gi
ng
s
ta
nda
r
d
a
s
s
e
s
s
m
e
nt
m
e
tr
ic
s
,
a
nd
th
e
f
unc
ti
ona
li
ty
o
f
th
e
a
lg
or
it
hm
is
c
om
pa
r
e
d
w
it
h
ot
h
e
r
s
ta
te
-
of
-
th
e
-
a
r
t
m
e
th
ods
.
T
he
a
lg
or
it
hm
is
s
im
ul
a
te
d
in
M
A
T
L
A
B
(
2022b)
s
of
twa
r
e
a
nd
a
n
N
V
I
D
I
A
gr
a
phi
c
s
c
a
r
d
e
qui
pp
e
d
w
it
h
8
G
ig
a
byt
e
s
of
onboa
r
d
R
A
M
.
I
n
tr
a
in
in
g
pha
s
e
,
w
e
a
ll
oc
a
te
70
%
of
th
e
da
t
a
.
S
pe
c
if
ic
a
ll
y,
7
,
011
da
ta
s
a
m
pl
e
s
a
r
e
f
e
d
in
to
th
e
pr
opos
e
d m
ode
l
a
s
t
r
a
in
in
g i
nput
, w
hi
le
3
,
004 da
ta
s
a
m
pl
e
s
a
r
e
s
e
t
a
s
id
e
f
or
t
he
t
e
s
ti
ng pha
s
e
. E
a
c
h i
m
a
ge
c
ont
r
ib
ut
e
s
1
,
000
f
e
a
tu
r
e
s
,
e
xt
r
a
c
te
d
by
th
e
pr
e
-
tr
a
in
e
d
c
onvolut
io
na
l
ne
twor
ks
.
S
ubs
e
que
nt
ly
,
th
e
M
R
M
R
a
lg
or
it
hm
i
s
e
m
pl
oye
d t
o r
e
duc
e
t
he
di
m
e
ns
io
ns
of
e
a
c
h f
e
a
tu
r
e
ve
c
to
r
t
o 500 dim
e
ns
io
ns
.
U
lt
i
m
a
te
ly
,
th
e
c
l
a
s
s
if
i
c
a
ti
on
of
th
e
s
e
f
e
a
t
ur
e
v
e
c
to
r
s
i
s
c
a
r
r
ie
d
o
ut
us
in
g
t
he
b
a
g
gi
n
g
m
e
t
ho
d.
F
ur
th
e
r
m
or
e
,
t
he
c
l
a
s
s
if
i
c
a
ti
o
n
a
c
c
ur
a
c
y
i
s
a
s
s
e
s
s
e
d
th
r
ou
gh
a
v
a
li
da
ti
o
n
pr
oc
e
s
s
,
i
nvo
lv
i
ng
t
h
e
p
a
r
ti
ti
o
ni
n
g
of
th
e
d
a
t
a
s
e
t
i
nt
o
t
e
n
f
ol
ds
.
T
he
m
o
de
l
i
s
it
e
r
a
ti
ve
l
y
tr
a
in
e
d
a
n
d
te
s
t
e
d
,
w
it
h
e
a
c
h
f
ol
d
s
e
r
vi
ng
a
s
a
v
a
li
da
ti
o
n
s
e
t.
N
ot
a
b
ly
,
th
e
d
iv
i
s
io
n
of
th
e
d
a
t
a
s
e
t
in
to
tr
a
in
i
ng
a
n
d
t
e
s
t
s
e
t
s
i
s
a
c
c
o
m
pl
is
h
e
d
t
hr
o
ugh
a
r
a
ndo
m
p
e
r
m
ut
a
ti
on
pr
o
c
e
s
s
.
T
h
e
f
i
nd
in
g
s
pr
e
s
e
nt
e
d
in
th
e
t
a
b
le
s
r
e
pr
e
s
e
n
t
t
he
a
ve
r
a
g
e
o
ut
c
om
e
of
50
-
ti
m
e
s
r
u
nn
in
g
t
h
e
pr
o
gr
a
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Sk
in
c
anc
e
r
di
agno
s
is
us
in
g
h
y
br
id
de
e
p p
r
e
-
tr
ai
ne
d c
on
v
ol
ut
io
nal
ne
ur
al
ne
tw
or
k
s
(
M
aha I
br
ahi
m
K
hal
e
e
l
)
2297
4
.1.
D
at
as
e
t
T
hi
s
w
or
k
ut
il
iz
e
s
th
e
H
AM
10000
da
ta
ba
s
e
f
or
th
e
pur
pos
e
of
di
a
gnos
in
g
s
ki
n
c
a
nc
e
r
.
T
he
H
A
M
10000
da
ta
s
e
t
c
om
pr
is
e
s
10,015
de
r
m
a
to
s
c
opy
pi
c
tu
r
e
s
,
e
a
c
h
e
xpos
in
g
di
s
ti
nc
t
f
or
m
s
of
s
ki
n
di
s
or
de
r
s
.
T
hi
s
da
t
a
ba
s
e
is
c
ol
le
c
te
d
f
r
om
A
us
tr
ia
a
nd
A
u
s
tr
a
li
a
pa
ti
e
nt
s
.
T
he
pi
c
tu
r
e
s
,
w
hi
c
h
a
r
e
c
e
nt
e
r
e
d
c
r
ops
,
ha
ve
di
m
e
ns
io
ns
of
600
×
450
pi
xe
l
s
.
T
he
da
ta
ba
s
e
i
s
c
a
te
gor
iz
e
d
in
to
7
c
la
s
s
e
s
,
w
it
h
e
a
c
h
c
la
s
s
s
ig
ni
f
yi
ng
a
di
s
ti
nc
t
di
s
e
a
s
e
. T
he
s
pe
c
if
ic
s
of
t
he
s
e
c
la
s
s
e
s
a
r
e
de
ta
il
e
d i
n T
a
bl
e
1.
S
e
ve
r
a
l
i
n
s
ta
nc
e
s
of
t
he
da
ta
s
e
t
pi
c
tu
r
e
s
a
r
e
di
s
pl
a
ye
d
in
F
ig
ur
e
6.
T
a
bl
e
1.
C
a
te
gor
ie
s
in
H
AM
10000
d
a
ta
s
e
t
D
i
s
e
a
s
e
t
ype
C
l
a
s
s
l
a
b
e
l
A
c
t
i
ni
c
ke
r
a
t
os
e
s
a
nd i
nt
r
a
e
pi
t
he
l
i
a
l
c
a
r
c
i
nom
a
/
B
ow
e
n di
s
e
a
s
e
(
a
ki
e
c
)
0
B
a
s
a
l
c
e
l
l
c
a
r
c
i
nom
a
(
bc
c
)
1
B
e
ni
gn l
e
s
i
ons
of
t
he
ke
r
a
t
os
i
s
(
bkl
)
2
D
e
r
m
a
t
of
i
br
om
a
(
df
)
3
M
e
l
a
nom
a
(
m
e
l
)
4
M
e
l
a
noc
yt
i
c
ne
vi
(
nv)
5
V
a
s
c
ul
a
r
l
e
s
i
on
s
(
va
s
c
)
6
F
ig
ur
e
6. D
a
ta
s
e
t
im
a
ge
e
xa
m
pl
e
s
4
.2.
E
val
u
at
io
n
c
r
it
e
r
ia
T
he
e
v
a
lu
a
ti
on
of
our
w
or
k
is
ba
s
e
d
on
f
our
ke
y
m
e
tr
ic
s
;
a
c
c
u
r
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
-
s
c
or
e
.
T
he
T
P
r
e
pr
e
s
e
nt
s
tr
ue
po
s
it
iv
e
de
te
c
ti
on
s
,
T
N
r
e
pr
e
s
e
nt
s
t
r
ue
ne
ga
ti
ve
de
te
c
ti
on
s
,
F
P
r
e
pr
e
s
e
nt
s
f
a
l
s
e
pos
it
iv
e
de
te
c
ti
ons
,
a
nd
F
N
r
e
pr
e
s
e
nt
s
f
a
ls
e
ne
ga
ti
ve
d
e
te
c
ti
ons
.
T
he
f
or
m
ul
a
s
f
or
th
e
s
e
m
e
tr
ic
s
a
r
e
pr
e
s
e
nt
e
d
in
(
7)
to
(
10
)
:
=
(
+
)
(
+
+
+
)
(
7)
=
(
+
)
(
8)
=
(
+
)
(
9)
1
=
2
∗
(
∗
)
(
+
)
(
10)
4
.3.
E
val
u
at
io
n
an
d
r
e
s
u
lt
s
c
om
p
ar
is
on
I
n
or
de
r
to
e
va
lu
a
te
th
e
s
y
s
te
m
’
s
e
f
f
e
c
ti
ve
n
e
s
s
in
de
te
c
ti
ng
s
ki
n
c
a
nc
e
r
,
w
e
ut
il
iz
e
d
a
c
onf
us
io
n
m
a
tr
ix
.
T
he
c
onf
u
s
io
n
m
a
tr
ix
f
or
th
e
te
s
t
da
t
a
is
pr
e
s
e
nt
e
d
in
il
l
us
tr
a
te
d
F
ig
ur
e
7. T
he
c
la
s
s
if
ic
a
ti
on
out
c
om
e
s
f
or
e
a
c
h
c
a
te
gor
y
a
s
pe
r
th
e
c
onf
us
io
n
m
a
tr
ix
a
r
e
di
s
pl
a
y
e
d.
T
he
ta
bl
e
s
ugge
s
ts
th
a
t
th
e
f
unc
ti
ona
li
ty
of
th
e
s
ugge
s
te
d a
ppr
oa
c
h i
s
de
e
m
e
d s
a
ti
s
f
a
c
to
r
y.
W
it
hi
n
th
e
r
e
a
lm
o
f
ML
,
th
e
e
f
f
ic
ie
nc
y
o
f
a
bi
na
r
y
c
la
s
s
if
ic
a
ti
on
a
t
di
f
f
e
r
e
nt
th
r
e
s
hol
ds
c
a
n
be
r
e
pr
e
s
e
nt
e
d
by
a
r
e
c
e
iv
e
r
ope
r
a
ti
ng
c
ha
r
a
c
te
r
is
ti
c
(
R
O
C
)
c
ur
ve
.
T
hi
s
c
ur
ve
is
c
on
s
tr
uc
te
d
us
in
g
two
ke
y
pa
r
a
m
e
te
r
s
:
th
e
tr
ue
po
s
it
iv
e
r
a
te
(
T
P
R
)
a
nd
th
e
f
a
l
s
e
po
s
it
iv
e
r
a
te
(
F
P
R
)
.
T
he
T
P
R
,
a
lt
e
r
na
ti
ve
ly
r
e
c
ogni
z
e
d
a
s
s
e
ns
it
iv
it
y
or
r
e
c
a
ll
,
de
not
e
s
th
e
f
r
a
c
ti
on
of
a
c
tu
a
l
pos
it
iv
e
s
a
m
pl
e
s
a
c
c
ur
a
te
ly
de
te
c
te
d.
I
n
c
ont
r
a
s
t,
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 3, J
une
2025
:
2291
-
2301
2298
F
P
R
r
e
pr
e
s
e
nt
s
th
e
f
r
a
c
ti
on
of
a
c
tu
a
l
ne
ga
ti
ve
s
a
m
pl
e
s
e
r
r
one
ous
ly
de
te
c
te
d
a
s
pos
it
iv
e
.
T
he
de
f
in
it
io
n
of
F
P
R
i
s
pr
e
s
e
nt
e
d
in
(
11)
.
=
+
(
11)
T
h
e
r
e
l
a
t
i
on
s
h
i
p
b
e
tw
e
e
n
a
c
l
a
s
s
i
f
i
e
r
’
s
T
P
R
a
n
d
F
P
R
i
s
d
e
p
i
c
t
e
d
b
y
g
r
a
p
hi
n
g
T
P
R
v
e
r
s
u
s
F
P
R
f
or
a
l
l
p
o
t
e
n
t
i
a
l
t
hr
e
s
h
o
l
d
s
.
A
h
i
gh
-
p
e
r
f
or
m
i
n
g
c
l
a
s
s
i
f
i
e
r
w
il
l
g
r
a
vi
t
a
t
e
t
o
w
a
r
d
s
t
h
e
u
pp
e
r
-
l
e
f
t
s
i
d
e
o
f
t
h
e
R
O
C
c
u
r
v
e
,
s
i
g
n
i
f
y
i
n
g
a
h
i
gh
T
P
R
a
n
d
a
m
in
i
m
a
l
F
P
R
.
I
n
c
o
n
tr
a
s
t
,
a
c
l
a
s
s
i
f
i
e
r
w
i
th
s
u
b
p
a
r
p
e
r
f
or
m
a
n
c
e
w
i
l
l
l
e
a
n
t
o
w
a
r
d
s
t
he
l
o
w
e
r
-
r
ig
h
t
c
o
r
n
e
r
,
i
n
di
c
a
t
in
g
a
d
im
i
n
i
s
h
e
d
T
P
R
a
n
d
a
n
e
l
e
v
a
te
d
F
P
R
.
A
c
l
a
s
s
i
f
i
e
r
t
h
a
t
p
e
r
f
or
m
s
a
t
a
l
e
v
e
l
e
q
u
i
v
a
l
e
nt
t
o
r
a
n
do
m
gu
e
s
s
i
n
g
w
il
l
c
oi
n
c
i
d
e
w
i
t
h
t
h
e
R
O
C
c
ur
v
e
’
s
d
i
a
g
o
n
a
l
l
in
e
,
d
e
n
ot
i
ng
a
n
e
q
u
a
l
T
P
R
a
n
d
F
P
R
.
F
ig
ur
e
7. C
onf
us
io
n m
a
tr
ix
of
s
ki
n c
a
nc
e
r
de
te
c
ti
on
T
he
R
O
C
c
ur
ve
f
or
th
e
pr
e
s
e
nt
e
d
m
e
th
od
is
il
lu
s
tr
a
te
d
in
F
ig
ur
e
8.
T
he
c
ur
ve
,
pos
it
io
ne
d
ne
a
r
th
e
to
p
-
le
f
t
c
or
ne
r
,
s
ig
ni
f
ie
s
a
not
a
bl
e
T
P
R
a
nd
a
c
or
r
e
s
pondingl
y
lo
w
F
P
R
.
T
hi
s
pos
it
io
ni
ng
in
di
c
a
te
s
th
a
t
ou
r
m
ode
l
is
hi
ghl
y
a
c
c
ur
a
te
in
de
te
c
ti
ng
s
ki
n
c
a
nc
e
r
.
F
ig
ur
e
9
il
lu
s
tr
a
te
s
a
ba
r
c
h
a
r
t
c
om
pa
r
is
on
de
pi
c
ti
ng
th
e
out
c
om
e
s
of
s
ki
n
c
a
nc
e
r
di
a
gno
s
is
u
s
in
g
our
pr
e
s
e
nt
e
d
a
ppr
oa
c
h
a
lo
ngs
id
e
v
a
r
io
us
ot
he
r
m
ode
ls
,
in
c
lu
di
ng
K
N
N
,
de
c
is
io
n
tr
e
e
s
(
D
T
)
,
S
V
M
,
C
N
N
,
a
nd
D
e
ns
e
N
e
t
.
A
s
de
p
ic
te
d
in
F
ig
ur
e
9,
th
e
K
N
N
,
D
T
,
S
V
M
,
C
N
N
,
a
nd
D
e
ns
e
N
e
t
m
ode
ls
a
c
hi
e
ve
d
a
c
c
ur
a
c
y
r
a
te
s
of
94.71,
95.14, 97.43,
92,
a
nd
83,
r
e
s
pe
c
ti
ve
ly
. T
he
pr
opos
e
d
m
e
th
od outpe
r
f
or
m
e
d t
he
s
e
m
ode
ls
w
it
h a
n i
m
pr
e
s
s
iv
e
a
c
c
ur
a
c
y r
a
te
of
99.
F
ig
ur
e
8. R
O
C
c
ur
ve
f
or
t
he
pr
e
s
e
nt
e
d a
ppr
oa
c
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
Sk
in
c
anc
e
r
di
agno
s
is
us
in
g
h
y
br
id
de
e
p p
r
e
-
tr
ai
ne
d c
on
v
ol
ut
io
nal
ne
ur
al
ne
tw
or
k
s
(
M
aha I
br
ahi
m
K
hal
e
e
l
)
2299
F
ig
ur
e
9.
T
he
e
f
f
ic
a
c
y of
t
he
s
ugge
s
te
d m
e
th
odol
ogy a
nd pr
e
-
tr
a
in
e
d m
e
th
ods
W
e
c
onduc
te
d a
c
om
pa
r
a
ti
ve
e
va
lu
a
ti
on of
ou
r
pr
opos
e
d a
ppr
o
a
c
h a
ga
in
s
t
va
r
io
us
ot
he
r
m
e
th
ods
f
o
r
s
ki
n
c
a
nc
e
r
de
te
c
ti
on.
A
s
pe
r
th
e
d
a
ta
in
T
a
bl
e
2,
our
m
o
de
l
a
c
hi
e
ve
d
a
hi
gh
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y
e
xc
e
e
di
ng
99%
.
T
hi
s
de
m
ons
tr
a
te
s
th
e
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
of
our
m
e
th
od
in
c
om
pa
r
is
on
to
ot
he
r
c
ont
e
m
por
a
r
y
a
lg
or
it
hm
s
.
O
ur
r
e
s
e
a
r
c
h
in
di
c
a
te
s
a
hi
ghe
r
r
e
c
a
l
l
doe
s
not
c
or
r
e
la
te
w
it
h
lo
w
e
r
p
r
e
c
is
io
n.
T
he
pr
e
s
e
nt
e
d
a
lg
or
it
hm
c
a
n
im
pr
ove
r
e
c
a
ll
m
e
tr
ic
w
it
hout
ne
ga
ti
v
e
ly
a
f
f
e
c
ti
ng
on
pr
e
c
is
io
n
c
r
it
e
r
io
n.
M
or
e
ove
r
,
w
e
di
s
c
ove
r
e
d
a
c
or
r
e
la
ti
on
be
twe
e
n
th
e
r
e
c
a
ll
a
nd
pr
e
c
is
io
n
m
e
tr
ic
s
w
it
h
th
e
F
-
s
c
or
e
c
r
i
te
r
io
n.
T
he
m
e
th
od
pr
e
s
e
nt
e
d i
n t
hi
s
r
e
s
e
a
r
c
h a
im
s
t
o a
c
hi
e
ve
a
s
ig
ni
f
ic
a
nt
ly
hi
ghe
r
F
-
s
c
or
e
r
a
ti
o c
om
pa
r
e
d t
o pr
e
vi
ous
s
tu
di
e
s
.
I
t
is
not
e
w
or
th
y w
hi
c
h t
hi
s
s
tu
dy ha
s
e
xa
m
in
e
d a
ll
t
he
m
e
tr
ic
s
o
f
p
r
e
c
is
io
n, F
-
s
c
or
e
,
a
c
c
ur
a
c
y,
r
e
c
a
ll
,
R
O
C
c
ur
ve
,
a
nd
le
a
r
ni
ng
c
ur
ve
s
im
ul
ta
ne
ous
ly
.
I
n
c
ont
r
a
s
t,
pr
io
r
s
tu
di
e
s
ha
ve
not
e
xpl
or
e
d
th
e
c
om
bi
ne
d
e
f
f
e
c
t
of
a
ll
th
e
s
e
c
r
it
e
r
ia
,
f
oc
us
in
g
in
s
te
a
d
on
onl
y
s
om
e
o
f
th
e
m
.
in
a
ddi
ti
on,
th
is
a
r
ti
c
le
is
pr
e
s
e
nt
in
g
a
c
om
pr
e
he
ns
iv
e
s
tu
dy
on
th
e
a
ut
om
a
ti
c
de
t
e
c
ti
on
of
s
ki
n
c
a
n
c
e
r
ba
s
e
d
on
tr
a
ns
f
e
r
le
a
r
ni
ng
us
in
g
a
hybr
id
m
e
th
od
in
c
lu
di
ng
th
r
e
e
pr
e
-
tr
a
in
e
d
ne
ur
a
l
ne
twor
ks
A
le
xN
e
t,
R
e
s
N
e
t5
0
,
a
nd
G
oogl
e
N
e
t
in
or
de
r
to
e
xt
r
a
c
t
f
e
a
tu
r
e
s
a
nd
ba
ggi
ng
te
c
hni
que
in
or
de
r
to
c
la
s
s
if
yi
ng.
H
ow
e
v
e
r
,
m
or
e
in
-
de
pt
h
r
e
s
e
a
r
c
he
s
c
a
n
be
a
ppl
ie
d
to
pr
ove
th
e
m
e
th
od
e
f
f
ic
ie
nc
y
on
a
la
r
ge
r
pa
ti
e
nt
popula
ti
on.
I
n
th
is
r
e
s
e
a
r
c
h,
it
w
a
s
di
s
pl
a
ye
d
th
a
t
m
a
th
e
m
a
ti
c
a
l
opt
im
iz
a
ti
on
m
e
th
ods
s
a
m
e
a
s
M
R
M
R
te
c
hni
que
a
r
e
m
or
e
e
f
f
ic
ie
nt
th
a
n
tr
a
di
ti
ona
l
a
lg
or
it
hm
s
f
or
s
e
le
c
ti
ng
opt
im
a
l
f
e
a
tu
r
e
s
.
F
ut
ur
e
w
or
ks
c
oul
d
e
xpl
or
e
o
th
e
r
nove
l
m
a
th
e
m
a
ti
c
a
l
m
e
th
od
s
in
or
de
r
to
opt
im
a
l
f
e
a
tu
r
e
s
s
e
le
c
ti
on.
R
e
c
e
nt
f
in
di
ngs
in
r
e
s
e
a
r
c
he
s
in
th
e
f
ie
ld
of
s
ki
n
c
a
nc
e
r
r
e
c
ogni
ti
on
di
s
pl
a
y
th
a
t
hybr
id
de
e
p
le
a
r
ni
ng
a
ppr
oa
c
he
s
c
a
n
a
c
hi
e
ve
hi
gh
pe
r
f
or
m
a
nc
e
in
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
f
r
om
im
a
ge
s
.
O
ur
obs
e
r
va
ti
on
di
s
pl
a
ys
c
onc
lu
s
iv
e
e
vi
de
nc
e
w
h
e
r
e
f
e
a
tu
r
e
e
xt
r
a
c
t
io
n
by
hybr
id
pr
e
-
tr
a
in
e
d
de
e
p
ne
ur
a
l
ne
twor
k
c
a
n
le
a
d
to
hi
ghe
r
e
f
f
e
c
ti
ve
ne
s
s
of
s
ki
n
c
a
nc
e
r
de
te
c
ti
on
s
ys
te
m
.
T
he
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
th
a
t
a
c
hi
e
ve
d
in
th
is
s
tu
dy pr
ove
s
t
hi
s
m
a
tt
e
r
.
T
a
bl
e
2.
R
e
s
ul
ts
c
om
p
a
r
is
on i
n t
e
r
m
s
of
e
va
lu
a
ti
on pa
r
a
m
e
te
r
s
M
e
t
hod
F
-
s
c
or
e
R
e
c
a
l
l
P
r
e
c
i
s
i
on
A
c
c
ur
a
c
y
KNN
[
28]
94.71
95.14
95.14
95
D
T
[
28]
95.14
95.57
95.71
95
S
V
M
[
28]
97.43
97.57
97.71
97
C
N
N
[
29]
92
89
91
93
D
e
ns
e
N
e
t
[
29]
83
84
93
95
T
he
pr
opos
e
d m
e
t
hod
98.88
98.90
98.91
98.9
5.
C
O
N
C
L
U
S
I
O
N
M
e
la
nom
a
,
s
ki
n
c
a
nc
e
r
kno
w
n
f
or
it
s
a
ggr
e
s
s
iv
e
m
e
ta
s
ta
ti
c
b
e
ha
vi
or
,
r
e
pr
e
s
e
nt
s
a
f
or
m
id
a
bl
e
a
nd
w
id
e
s
pr
e
a
d
m
a
li
gna
nc
y
w
it
hi
n
th
e
s
pe
c
tr
um
of
s
ki
n
-
r
e
la
te
d
c
a
nc
e
r
s
.
P
r
om
pt
a
nd
a
c
c
ur
a
te
di
a
gnos
is
of
m
e
la
nom
a
is
c
r
uc
ia
l
in
e
nha
nc
in
g
th
e
c
ha
nc
e
s
of
e
f
f
e
c
ti
ve
tr
e
a
tm
e
nt
.
T
he
m
a
jo
r
f
oc
us
of
th
is
r
e
s
e
a
r
c
h
is
to
put
f
or
w
a
r
d
a
s
tu
r
dy
m
ode
l
to
s
ki
n
c
a
nc
e
r
r
e
c
ogni
ti
on
,
ut
il
iz
in
g
th
e
pot
e
nt
ia
l
of
pr
e
-
e
xi
s
ti
ng
de
e
p
ne
ur
a
l
ne
twor
ks
.
T
he
in
he
r
e
nt
de
pt
h
a
nd
e
f
f
e
c
ti
ve
ne
s
s
of
c
onvolut
io
n
f
il
te
r
s
r
e
nde
r
pr
e
-
tr
a
in
e
d
C
N
N
s
pa
r
ti
c
ul
a
r
l
y
a
dva
nt
a
ge
ous
f
or
im
a
ge
-
r
e
la
te
d
ta
s
ks
.
D
e
e
p
a
r
c
hi
te
c
tu
r
e
s
,
in
c
lu
di
ng
R
e
s
N
e
t,
G
oogl
e
N
e
t
,
a
nd
A
le
xN
e
t
a
r
e
e
m
pl
oye
d
f
or
e
xt
r
a
c
ti
on
of
f
e
a
tu
r
e
s
.
F
ol
lo
w
in
g
th
e
e
xt
r
a
c
ti
on
a
nd
in
te
gr
a
ti
on
of
f
e
a
tu
r
e
s
,
th
e
M
R
M
R
a
lg
or
it
hm
is
a
ppl
ie
d
in
th
e
m
e
th
odol
ogy
to
s
e
le
c
t
th
e
m
os
t
pe
r
t
in
e
nt
f
e
a
tu
r
e
s
.
A
ddi
ti
ona
ll
y,
th
e
c
la
s
s
if
ic
a
ti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 3, J
une
2025
:
2291
-
2301
2300
pha
s
e
in
c
or
por
a
te
s
th
e
ba
ggi
ng
a
lg
or
it
hm
to
e
nha
nc
e
th
e
m
ode
l’
s
r
obus
tn
e
s
s
.
A
n
e
xha
us
ti
ve
c
om
pa
r
is
on
w
a
s
c
onduc
te
d,
w
he
r
e
in
our
pr
opos
e
d
m
e
th
odol
ogy
w
a
s
ju
xt
a
pos
e
d
a
ga
in
s
t
va
r
io
us
a
lt
e
r
na
ti
ve
s
in
c
lu
di
ng
K
N
N
,
D
T
,
S
V
M
,
C
N
N
,
a
nd
D
e
n
s
e
N
e
t
in
th
e
dom
a
in
of
s
ki
n
c
a
nc
e
r
r
e
c
ogni
ti
on
.
T
he
r
e
s
ul
ts
und
e
r
s
c
or
e
th
e
out
s
ta
ndi
ng pe
r
f
or
m
a
nc
e
of
our
a
ppr
oa
c
h, s
ur
pa
s
s
in
g t
ha
t
of
ot
he
r
c
ont
e
m
por
a
r
y a
lg
or
it
hm
s
i
n t
he
f
ie
ld
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
A
ut
hor
s
s
ta
te
no f
undi
ng i
nvol
ve
d.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on
.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
M
a
ha
I
br
a
hi
m
K
ha
le
e
l
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
A
m
ir
L
a
ki
z
a
de
h
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
A
ut
hor
s
s
ta
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
da
ta
th
a
t
s
uppor
t
th
e
f
in
di
ngs
of
th
i
s
s
tu
dy
a
r
e
ope
nl
y
a
va
il
a
bl
e
in
[
K
a
ggl
e
]
a
t
ht
tp
s
:/
/ww
w
.ka
ggl
e
.c
om
/d
a
ta
s
e
ts
/k
m
a
de
r
/s
ki
n
-
c
a
n
c
e
r
-
m
ni
s
t
-
ha
m
10000
.
R
E
F
E
R
E
N
C
E
S
[
1]
M
.
C
i
ą
ż
yńs
ka
e
t
al
.
,
“
T
he
i
nc
i
de
nc
e
a
nd
c
l
i
ni
c
a
l
a
n
a
l
ys
i
s
of
non
-
m
e
l
a
nom
a
s
k
i
n
c
a
nc
e
r
,”
S
c
i
e
nt
i
f
i
c
R
e
por
t
s
,
vol
.
11,
no.
1,
2021,
doi
:
10.1038/
s
41598
-
021
-
83502
-
8.
[
2]
E
.
R
.
P
a
r
ke
r
,
“
T
he
i
nf
l
ue
nc
e
of
c
l
i
m
a
t
e
c
ha
nge
on
s
ki
n
c
a
nc
e
r
i
nc
i
de
nc
e
–
a
r
e
vi
e
w
of
t
he
e
vi
de
nc
e
,
”
I
nt
e
r
nat
i
onal
J
ou
r
nal
of
W
om
e
n’
s
D
e
r
m
at
ol
ogy
, vol
. 7, no. 1, pp. 17
–
27, J
a
n. 2021, doi
:
10.1016/
j
.i
j
w
d.2020.07.003.
[
3]
L
.
R
e
y
-
B
a
r
r
os
o,
S
.
P
e
ña
-
G
ut
i
é
r
r
e
z
,
C
.
Y
á
ñe
z
,
F
.
J
.
B
ur
gos
-
F
e
r
ná
nde
z
,
M
.
V
i
l
a
s
e
c
a
,
a
nd
S
.
R
oyo,
“
O
pt
i
c
a
l
t
e
c
hnol
ogi
e
s
f
or
t
he
i
m
pr
ove
m
e
nt
of
s
ki
n c
a
nc
e
r
di
a
gnos
i
s
:
a
r
e
v
i
e
w
,”
S
e
ns
or
s
, vol
. 21, no. 1, J
a
n. 2021, doi
:
10.3390/
s
21010252.
[
4]
M
.
A
.
K
ha
n,
M
.
S
ha
r
i
f
,
T
.
A
kr
a
m
,
S
.
K
a
dr
y,
a
nd
C
.
H
s
u,
“
A
t
w
o‐
s
t
r
e
a
m
de
e
p
ne
ur
a
l
ne
t
w
or
k‐
ba
s
e
d
i
nt
e
l
l
i
ge
nt
s
ys
t
e
m
f
or
c
om
pl
e
x
s
ki
n
c
a
nc
e
r
t
yp
e
s
c
l
a
s
s
i
f
i
c
a
t
i
on,”
I
nt
e
r
nat
i
onal
J
our
nal
of
I
nt
e
l
l
i
ge
nt
Sy
s
t
e
m
s
,
vol
.
37,
no.
12,
pp.
10621
–
10649,
2022
,
doi
:
10.1002/
i
nt
.22691.
[
5]
P
.
F
ont
a
ni
l
l
a
s
e
t
al
.
,
“
D
i
s
e
a
s
e
r
i
s
k
s
c
or
e
s
f
or
s
ki
n
c
a
nc
e
r
s
,”
N
at
ur
e
C
om
m
uni
c
at
i
ons
,
vol
.
12,
no.
1,
J
a
n.
2021,
doi
:
10.1038/
s
41467
-
020
-
20246
-
5.
[
6]
M
.
A
kt
e
r
e
t
al
.
,
“
A
n
i
nt
e
gr
a
t
e
d
de
e
p
l
e
a
r
ni
ng
m
ode
l
f
or
s
ki
n
c
a
nc
e
r
d
e
t
e
c
t
i
on
us
i
ng
hybr
i
d
f
e
a
t
ur
e
f
us
i
on
t
e
c
hni
q
ue
,”
B
i
om
e
di
c
al
M
at
e
r
i
al
s
&
D
e
v
i
c
e
s
, J
a
n. 2025, doi
:
10.1007/
s
44174
-
024
-
00264
-
3.
[
7]
M
.
D
i
l
da
r
e
t
al
.
,
“
S
ki
n
c
a
nc
e
r
de
t
e
c
t
i
on:
a
r
e
vi
e
w
us
i
ng
de
e
p
l
e
a
r
ni
ng
t
e
c
hni
qu
e
s
,”
I
nt
e
r
nat
i
onal
J
ou
r
nal
of
E
nv
i
r
on
m
e
nt
al
R
e
s
e
ar
c
h and P
ubl
i
c
H
e
al
t
h
, vol
. 18, no. 10, M
a
y 2021, doi
:
10.3390/
i
j
e
r
ph181
05479.
[
8]
H
.
B
ha
t
t
,
V
.
S
ha
h,
K
.
S
ha
h,
R
.
S
ha
h,
a
nd
M
.
S
ha
h,
“
S
t
a
t
e
-
of
-
t
he
-
a
r
t
m
a
c
hi
ne
l
e
a
r
ni
ng
t
e
c
hni
que
s
f
or
m
e
l
a
nom
a
s
ki
n
c
a
n
c
e
r
de
t
e
c
t
i
on
a
nd
c
l
a
s
s
i
f
i
c
a
t
i
on:
a
c
om
pr
e
he
ns
i
ve
r
e
vi
e
w
,”
I
nt
e
l
l
i
ge
nt
M
e
di
c
i
ne
,
vol
.
3,
no.
3,
pp.
180
–
190,
A
ug.
2023,
doi
:
10.1016/
j
.i
m
e
d.2022.08.004.
[
9]
T
.
M
a
z
ha
r
e
t
al
.
,
“
T
he
r
ol
e
o
f
m
a
c
hi
ne
l
e
a
r
ni
ng
a
nd
de
e
p
l
e
a
r
ni
ng
a
ppr
oa
c
he
s
f
or
t
he
de
t
e
c
t
i
on
of
s
ki
n
c
a
nc
e
r
,”
H
e
al
t
hc
ar
e
,
vol
. 11, no. 3, F
e
b. 2023, doi
:
10.3390/
he
a
l
t
hc
a
r
e
11030415.
[
10]
V
.
A
.
O
.
N
a
nc
y,
P
.
P
r
a
bha
va
t
hy,
M
.
S
.
A
r
ya
,
a
nd
B
.
S
.
A
ha
m
e
d,
“
C
om
pa
r
a
t
i
ve
s
t
udy
a
nd
a
na
l
ys
i
s
on
s
ki
n
c
a
n
c
e
r
de
t
e
c
t
i
on
us
i
n
g
m
a
c
hi
ne
l
e
a
r
ni
ng
a
nd
de
e
p
l
e
a
r
ni
ng
a
l
gor
i
t
hm
s
,”
M
ul
t
i
m
e
di
a
T
ool
s
and
A
ppl
i
c
at
i
ons
,
vol
.
82,
no.
29,
pp.
45913
–
45957,
D
e
c
.
2023, doi
:
10.1007/
s
11042
-
023
-
16422
-
6.
[
11]
L
.
S
t
e
e
l
e
,
X
.
L
.
T
a
n,
B
.
O
l
a
bi
,
J
.
M
.
G
a
o,
R
.
J
.
T
a
na
ka
,
a
nd
H
.
C
.
W
i
l
l
i
a
m
s
,
“
D
e
t
e
r
m
i
ni
ng
t
he
c
l
i
ni
c
a
l
a
ppl
i
c
a
bi
l
i
t
y
of
m
a
c
hi
n
e
l
e
a
r
ni
ng m
ode
l
s
t
hr
ough a
s
s
e
s
s
m
e
nt
of
r
e
por
t
i
ng a
c
r
os
s
s
ki
n phot
ot
ype
s
a
nd r
a
r
e
r
s
ki
n c
a
nc
e
r
t
ype
s
:
a
s
y
s
t
e
m
a
t
i
c
r
e
vi
e
w
,”
J
our
na
l
of
t
he
E
ur
ope
an
A
c
ade
m
y
of
D
e
r
m
at
ol
ogy
and
V
e
ne
r
e
ol
ogy
,
vo
l
.
37,
no.
4,
pp.
657
–
665,
A
pr
.
2023,
doi
:
10.1111/
j
dv.18814.
[
12]
M
.
Z
a
f
a
r
,
M
.
S
ha
r
i
f
,
M
.
S
ha
r
i
f
,
S
.
K
a
dr
y,
S
.
B
ukha
r
i
,
a
nd
H
.
R
a
uf
,
“
S
ki
n
l
e
s
i
on
a
na
l
ys
i
s
a
nd
c
a
nc
e
r
de
t
e
c
t
i
on
ba
s
e
d
on
m
a
c
hi
ne
/
de
e
p l
e
a
r
ni
ng t
e
c
hni
que
s
:
a
c
om
pr
e
he
ns
i
ve
s
ur
ve
y,
”
L
i
f
e
, vol
. 13, no. 1, J
a
n. 2023, doi
:
10.3390/
l
i
f
e
13010146.
[
13]
R
.
L
ou,
Z
.
L
v,
S
.
D
a
ng,
T
.
S
u,
a
nd
X
.
L
i
,
“
A
ppl
i
c
a
t
i
on
of
m
a
c
hi
ne
l
e
a
r
ni
ng
i
n
oc
e
a
n
da
t
a
,
”
M
ul
t
i
m
e
di
a
Sy
s
t
e
m
s
,
vol
.
29,
no.
3
,
pp. 1815
–
1824, J
un. 2023, doi
:
10.1007/
s
00530
-
020
-
00733
-
x.
Evaluation Warning : The document was created with Spire.PDF for Python.