I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
2
,
A
p
r
il 2
0
2
6
,
p
p
.
1
5
2
9
~
1
5
3
8
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
2
.
p
p
1
5
2
9
-
1
5
3
8
1529
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
2D
-
CNN
-
G
ACL
-
ECGNe
t
g
ra
ph at
tent
io
n:
a ro
bust
framewo
rk
for
elec
troca
rdio
g
ra
m
-
ba
sed stress
detec
tion
P.
K
a
v
it
ha
,
L
.
Sh
a
k
k
ee
r
a
P
r
e
si
d
e
n
c
y
S
c
h
o
o
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
P
r
e
s
i
d
e
n
c
y
U
n
i
v
e
r
si
t
y
,
B
e
n
g
a
l
u
r
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
J
u
n
2
,
2
0
2
5
R
ev
is
ed
Dec
2
7
,
2
0
2
5
Acc
ep
ted
J
an
2
5
,
2
0
2
6
Early
d
e
tec
ti
o
n
o
f
c
a
rd
io
v
a
sc
u
lar
d
ise
a
se
s
(CVD
s)
v
ia
e
lec
tro
c
a
rd
io
g
ra
m
(ECG
)
c
las
sific
a
ti
o
n
d
u
ri
n
g
p
h
y
sio
l
o
g
ica
l
stre
ss
is
c
rit
ica
l
a
n
d
re
m
a
in
s
c
h
a
ll
e
n
g
i
n
g
d
u
e
to
stre
ss
-
in
d
u
c
e
d
m
o
rp
h
o
lo
g
ica
l
v
a
riab
i
li
ty
,
n
o
ise
fro
m
a
m
b
u
lato
ry
se
tt
i
n
g
s
,
a
n
d
in
ter
-
c
l
a
ss
a
m
b
ig
u
it
ies
.
Ex
isti
n
g
m
o
d
e
ls
,
su
c
h
a
s
1
D
sig
n
a
l
-
b
a
se
d
m
o
d
e
ls
wi
th
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
s
(C
NN
s)
a
n
d
g
ra
p
h
c
o
n
v
o
lu
t
io
n
a
l
n
e
tw
o
rk
s
(G
CNs
),
stru
g
g
le
to
a
d
a
p
t
t
o
d
y
n
a
m
ic
stre
ss
c
o
n
d
i
ti
o
n
s
a
n
d
g
e
n
e
ra
te
in
terp
re
t
a
b
le
in
sig
h
ts.
I
n
re
sp
o
n
se
,
we
p
r
o
p
o
se
2
D
-
CNN
a
n
d
g
ra
p
h
a
tt
e
n
ti
o
n
n
e
tw
o
rk
(G
AT)
fo
r
o
p
ti
m
ize
r.
T
h
e
m
o
d
e
l
2
D
-
CNN
a
n
d
G
ACL
-
ECG
-
Ne
t,
a
n
in
n
o
v
a
ti
v
e
fra
m
e
wo
rk
i
n
teg
ra
ti
n
g
G
ATs
with
a
d
a
p
ti
v
e
c
o
n
tras
ti
v
e
lea
rn
i
n
g
(ACL)
a
n
d
m
o
rp
h
o
-
tem
p
o
r
a
l
g
ra
p
h
c
o
n
stru
c
ti
o
n
.
Ke
y
i
n
n
o
v
a
ti
o
n
s
in
c
lu
d
e
2D
-
CNN
d
e
n
o
isi
n
g
,
2
D
tran
sfo
rm
a
ti
o
n
,
d
y
n
a
m
ic
m
o
r
p
h
o
-
tem
p
o
ra
l
g
ra
p
h
s
m
o
d
e
li
n
g
ECG
b
e
a
ts
a
s
n
o
d
e
s
with
h
y
b
rid
e
d
g
e
s
(7
0
%
m
o
rp
h
o
lo
g
ica
l
sim
il
a
rit
y
,
3
0
%
tem
p
o
ra
l
p
ro
x
imity
),
a
n
d
stre
ss
-
a
d
a
p
ti
v
e
c
o
n
tras
ti
v
e
lo
ss
with
lea
rn
a
b
le
m
a
rg
in
s
o
n
stre
ss
-
c
o
n
d
it
i
o
n
e
d
lab
e
ls,
re
d
u
c
i
n
g
c
las
s
a
m
b
ig
u
it
y
b
y
1
8
%
.
M
u
lt
i
-
h
e
a
d
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
s
p
ro
v
id
e
in
terp
re
tab
le
h
e
a
tma
p
s
a
li
g
n
e
d
wit
h
c
a
rd
io
lo
g
ist
a
n
n
o
tati
o
n
s
(κ
=
0
.
8
2
)
a
n
d
a
re
e
v
a
lu
a
te
d
u
si
n
g
M
a
ss
a
c
h
u
se
tt
s
In
stit
u
te
o
f
Tec
h
n
o
lo
g
y
-
Be
th
Isra
e
l
Ho
sp
it
a
l
(M
IT
-
BIH)
a
rrh
y
th
m
i
a
d
a
tab
a
se
,
we
a
ra
b
le
stre
ss
a
n
d
a
ffe
c
t
d
e
tec
ti
o
n
(W
ES
AD
)
d
a
tas
e
t
fo
r
e
m
o
ti
o
n
a
l
stre
ss
,
a
n
d
stre
ss
a
t
wo
rk
,
k
n
o
wle
d
g
e
wo
r
k
(
S
WE
LL
-
K
W)
d
a
tas
e
t
fo
r
c
o
g
n
i
ti
v
e
stre
ss
.
2
D
-
CNN
-
G
ACL
-
EC
G
-
N
e
t
a
c
h
iev
e
s
sta
te
-
of
-
th
e
-
a
rt
p
e
rfo
rm
a
n
c
e
with
9
8
.
7
%
F
1
-
sc
o
r
e
(M
IT
-
BIH),
9
4
.
2
%
(W
ES
AD
),
a
n
d
9
2
.
8
%
(S
WE
LL
-
KW)
,
o
u
t
p
e
rfo
rm
i
n
g
C
NN
-
b
id
irec
ti
o
n
a
l
lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
(BiL
S
TM
)
a
n
d
G
CN
b
a
se
li
n
e
s
b
y
9
5
%
.
T
h
e
fra
m
e
wo
rk
is
c
o
m
p
u
tatio
n
a
ll
y
e
fficie
n
t
a
n
d
c
li
n
ica
ll
y
v
a
li
d
a
ted
f
o
r
we
a
ra
b
le h
e
a
lt
h
m
o
n
it
o
rin
g
.
K
ey
w
o
r
d
s
:
Ad
ap
tiv
e
co
n
tr
asti
v
e
lear
n
in
g
E
C
G
clas
s
if
icatio
n
Fu
s
io
n
b
ea
ts
Gr
ap
h
n
e
u
r
al
n
etwo
r
k
Stre
s
s
-
awa
r
en
ess
lear
n
in
g
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
:
P.
Kav
ith
a
Pre
s
id
en
cy
Sch
o
o
l o
f
C
o
m
p
u
t
er
Scien
ce
an
d
E
n
g
in
ee
r
i
n
g
,
P
r
esid
en
cy
Un
iv
er
s
ity
B
en
g
alu
r
u
,
Kar
n
atak
a,
I
n
d
ia
E
m
ail:
k
av
ith
a.
2
0
2
3
3
C
SE0
0
2
4
@
p
r
esid
en
cy
u
n
iv
e
r
s
ity
.
in
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
W
o
r
l
d
H
e
a
l
th
Or
g
an
i
z
a
ti
o
n
(
W
H
O
)
r
e
p
o
r
t
s
s
t
a
t
e
th
a
t
c
a
r
d
i
o
v
a
s
c
u
la
r
d
i
s
e
a
s
e
s
(
C
V
D
s
)
c
au
s
e
d
e
a
t
h
s
o
f
1
7
.
9
m
i
l
l
i
o
n
an
n
u
al
l
y
w
o
r
l
d
w
id
e
[
1
]
.
A
c
u
te
s
t
r
e
s
s
t
e
n
d
s
t
o
w
o
r
s
en
co
n
d
i
t
i
o
n
s
l
i
k
e
m
y
o
c
a
r
d
i
a
l
i
s
c
h
e
m
i
a
an
d
a
r
r
h
y
t
h
m
i
a
s
,
in
c
l
u
d
i
n
g
p
r
em
a
t
u
r
e
v
en
t
r
ic
u
l
a
r
co
n
tr
a
c
t
io
n
s
(
P
V
C
s
)
a
n
d
a
t
r
i
al
p
r
em
a
t
u
r
e
c
o
n
tr
a
c
t
i
o
n
s
(
A
P
C
s
)
.
A
d
d
i
t
i
o
n
a
l
l
y
,
co
n
d
u
c
t
io
n
a
b
n
o
r
m
a
l
i
ti
e
s
s
u
c
h
a
s
r
ig
h
t
b
u
n
d
l
e
b
r
an
c
h
b
lo
c
k
(
R
B
B
B
)
a
n
d
l
ef
t
b
u
n
d
l
e
b
r
an
c
h
b
l
o
c
k
(
L
B
B
B
)
c
a
n
a
f
f
e
c
t
th
e
h
e
a
r
t
s
i
g
n
a
l
s
[
1
]
.
U
s
in
g
an
ad
a
p
t
iv
e
n
o
i
s
e
t
e
ch
n
iq
u
e
t
h
e
c
o
m
p
l
ex
s
i
g
n
a
l
s
a
r
e
s
p
l
i
t
s
i
n
t
o
i
n
tr
i
n
s
i
c
m
o
d
e
f
u
n
c
t
i
o
n
s
i
n
co
m
p
l
e
t
e
en
s
e
m
b
le
e
m
p
ir
i
c
a
l
m
o
d
e
d
e
c
o
m
p
o
s
i
t
i
o
n
w
i
t
h
ad
a
p
t
iv
e
n
o
i
s
e
(
C
E
E
M
D
A
N)
f
o
r
a
c
cu
r
a
cy
.
T
h
e
s
y
n
t
h
e
t
ic
s
a
m
p
l
e
s
a
r
e
g
en
e
r
a
t
e
d
f
o
r
m
i
n
o
r
i
t
y
c
l
a
s
s
e
s
t
o
b
a
l
a
n
c
e
d
a
t
a
s
e
t
s
i
n
s
y
n
t
h
e
t
i
c
m
in
o
r
i
ty
o
v
er
-
s
a
m
p
l
i
n
g
t
e
c
h
n
i
q
u
e
(
SM
O
T
E
)
.
T
h
e
g
r
a
p
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
5
2
9
-
1
5
3
8
1530
n
e
u
r
a
l
n
e
t
w
o
r
k
(
G
N
N
)
i
n
g
r
a
p
h
a
t
t
en
t
i
o
n
n
e
t
wo
r
k
(
G
AT
)
u
s
e
s
a
t
t
en
t
i
o
n
to
l
e
ar
n
n
o
d
e
r
e
l
a
t
io
n
s
h
i
p
s
.
F
e
a
t
u
r
e
i
m
p
r
o
v
em
e
n
t
in
a
d
a
p
t
i
v
e
co
n
tr
a
s
t
i
v
e
l
ea
r
n
in
g
(
A
C
L
)
th
r
o
u
g
h
t
h
e
co
m
p
a
r
i
s
o
n
o
f
s
im
i
l
a
r
an
d
d
i
s
s
i
m
i
l
a
r
s
am
p
l
e
s
.
T
h
e
h
e
a
r
t
r
a
t
e
v
ar
i
a
b
i
l
i
ty
(
H
R
V
)
m
ea
s
u
r
e
s
t
i
m
e
i
n
t
er
v
a
l
s
i
n
h
e
ar
t
b
e
a
t
s
a
n
d
m
ir
r
o
r
in
g
a
u
t
o
n
o
m
i
c
n
er
v
o
u
s
s
y
s
t
e
m
ac
t
i
v
i
t
y
.
A
b
n
o
r
m
a
l
c
a
r
d
i
a
c
r
h
y
t
h
m
i
s
P
V
C
,
A
P
C
,
R
B
B
B
,
a
n
d
L
B
B
B
w
i
t
h
i
r
r
eg
u
l
ar
co
n
d
u
c
t
io
n
p
a
t
t
er
n
s
.
T
r
ad
itio
n
al
E
C
G
class
if
ier
s
ten
d
to
f
ail
u
n
d
e
r
s
tr
ess
-
in
d
u
ce
d
v
ar
iab
ilit
y
,
s
u
ch
as
S
T
-
s
eg
m
en
t
d
ep
r
ess
io
n
d
u
r
in
g
m
en
tal
s
tr
es
s
o
r
HR
V
s
h
if
ts
in
co
g
n
itiv
e
ta
s
k
s
,
lead
in
g
to
cr
itical
m
is
d
iag
n
o
s
es
[
2
]
,
[
3
]
.
Fo
r
ex
am
p
le,
s
tr
ess
-
in
d
u
ce
d
s
in
u
s
tach
y
ca
r
d
ia
m
im
ics
atr
ial
f
ib
r
illatio
n
,
wh
ile
m
o
tio
n
ar
tifa
cts
in
wea
r
ab
le
E
C
Gs
m
ay
o
b
s
cu
r
e
PVC
s
o
r
tr
an
s
ien
t
R
B
B
B
/L
B
B
B
p
atter
n
s
,
r
is
k
in
g
in
ap
p
r
o
p
r
iate
tr
ea
tm
en
t
[
4
]
,
[
5
]
.
Hea
r
tb
ea
ts
o
r
ca
r
d
iac
r
h
y
th
m
ca
n
b
e
class
if
ied
u
s
in
g
s
ev
er
al
ap
p
r
o
ac
h
es
lik
e
n
o
r
m
al
b
ea
ts
,
s
u
p
r
av
en
tr
icu
lar
b
ea
ts
,
v
en
tr
icu
lar
b
ea
ts
,
an
d
f
u
s
io
n
b
ea
ts
(
N,
S,
V,
an
d
F)
in
clu
d
in
g
m
a
n
u
al
in
ter
p
r
etatio
n
o
f
elec
tr
o
ca
r
d
io
g
r
am
(
E
C
G)
tr
ac
in
g
s
b
y
ex
p
er
ie
n
c
ed
an
d
tr
ain
e
d
h
ea
lth
ca
r
e
wo
r
k
er
s
o
r
au
to
m
ated
class
if
icatio
n
u
s
in
g
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
ith
m
s
f
o
r
e
n
h
an
ce
d
ac
cu
r
ac
y
.
E
C
G
tr
ac
in
g
ex
h
ib
it
ch
ar
ac
ter
is
tics
wav
es
an
d
co
m
p
lex
es,
wh
ich
id
en
tifie
d
b
y
th
eir
s
h
ap
e
a
n
d
d
u
r
atio
n
in
h
ea
r
t
b
ea
ts
b
y
o
u
r
b
o
th
a
p
p
r
o
ac
h
es,
wh
ich
f
ea
tu
r
e
ex
tr
ac
tio
n
ar
e
g
u
id
ed
b
y
ad
v
an
ce
m
e
n
t
in
m
ed
ical
in
s
tr
u
m
en
tatio
n
E
C
5
7
s
tan
d
ar
d
,
wh
ich
ar
e
ab
le
to
d
i
s
tin
g
u
is
h
h
ea
r
tb
ea
ts
in
to
v
ar
i
o
u
s
ca
teg
o
r
ies
s
u
ch
as
PVC
s
,
AP
C
s
,
R
B
B
B
,
an
d
L
B
B
B
.
Fig
u
r
e
1
s
h
o
w
s
th
e
E
C
G
s
ig
n
al
class
if
icatio
n
o
f
a
h
ea
lth
y
h
e
ar
t.
I
t
co
n
s
is
ts
o
f
th
r
ee
m
ajo
r
p
ar
ts
o
f
wa
v
co
m
p
o
n
e
n
ts
,
th
e
P
wav
e,
th
e
QR
S
co
m
p
lex
,
an
d
th
e
T
wav
e.
T
h
e
v
ital sig
n
s
o
f
E
C
G
clas
s
if
icati
o
n
will d
is
p
lay
s
in
clu
d
in
g
PQ,
ST,
QR
S,
an
d
QT
in
ter
v
als ar
e
cr
u
cial
co
m
p
o
n
e
n
ts
in
d
icato
r
f
o
r
d
e
v
elo
p
in
g
E
C
G
clas
s
if
icatio
n
m
o
d
els.
Fig
u
r
e
1
.
E
C
G
class
if
icatio
n
Ho
wev
er
,
th
eir
f
ix
e
d
r
esp
ec
tiv
e
f
ield
s
in
c
o
n
v
o
lu
tio
n
al
n
eu
r
a
l
n
etwo
r
k
s
(
C
NNs)
an
d
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs
)
,
lik
e
R
esNe
t
-
3
4
,
p
er
f
o
r
m
s
well
(
9
5
%
ac
cu
r
ac
y
o
n
Ma
s
s
ac
h
u
s
etts
I
n
s
tit
u
te
o
f
T
ec
h
n
o
lo
g
y
-
B
eth
I
s
r
ae
l
Ho
s
p
ital
(
MI
T
-
B
I
H)
)
wh
ich
ca
u
s
es
m
alf
u
n
ctio
n
u
n
d
er
s
tr
ess
ac
tiv
ity
[
6
]
.
S
tr
ess
ad
ap
tatio
n
is
lim
ited
b
y
GNNs,
wh
ich
e
m
p
lo
y
s
tatic
g
r
a
p
h
s
b
u
t
d
is
r
eg
ar
d
tem
p
o
r
al
d
y
n
am
ics
[
7
]
.
Ad
d
itio
n
ally
,
o
v
er
lap
s
b
r
o
u
g
h
t
o
n
b
y
s
tr
ess
ar
e
p
r
o
b
l
em
atic
f
o
r
f
ix
ed
-
m
ar
g
in
tr
i
p
let
lo
s
s
(
e.
g
.
,
2
5
%
f
alse
p
o
s
itiv
es
b
etwe
en
an
x
iety
an
d
atr
ial
f
ib
r
illatio
n
)
.
T
h
is
o
v
er
lap
is
r
ed
u
ce
d
b
y
3
2
%
b
y
o
u
r
ad
ap
ti
v
e
m
ar
g
in
s
,
wh
ich
ar
e
lear
n
ed
b
y
s
tr
ess
lab
elin
g
(
m
u
lti
-
lay
er
p
er
ce
p
tr
o
n
(
ML
P)
with
r
ec
tifie
d
li
n
ea
r
u
n
it
(
R
eL
U
)
)
[
8
]
.
Ou
r
GAT
s
,
HR
V
,
an
d
ST
-
s
eg
m
en
ts
d
u
r
in
g
h
ig
h
s
tr
ess
,
wh
ile
p
r
ev
io
u
s
r
esear
ch
ap
p
ly
u
n
if
o
r
m
atten
tio
n
wi
th
o
u
t
s
tr
ess
-
awa
r
e
m
ec
h
an
is
m
s
[
9
]
.
Stre
s
s
-
in
d
u
c
ed
E
C
G
v
ar
ia
b
ilit
y
in
d
y
n
am
ic
g
r
ap
h
c
r
ea
tio
n
with
ad
a
p
tiv
e
ed
g
e
weig
h
tin
g
m
o
d
el
will e
f
f
icien
tly
ca
teg
o
r
i
ze
[
1
0
]
u
s
in
g
its
ST
-
s
eg
m
en
t sh
if
ts
an
d
h
ea
r
t
r
ate.
L
ik
e
co
n
tr
asti
v
e
lear
n
in
g
o
f
ca
r
d
iac
s
ig
n
als
(
C
L
OC
S
)
i
n
co
n
tr
asti
v
e
lear
n
in
g
f
r
am
e
wo
r
k
s
to
d
if
f
er
en
tiate
ca
r
d
iac
s
ig
n
als
b
etwe
en
in
d
iv
id
u
al
p
er
s
o
n
a
n
d
co
n
te
x
t
ar
e
ap
p
ly
in
s
tr
ess
-
ad
ap
tiv
e
m
ar
g
i
n
s
.
T
h
ese
tech
n
iq
u
es,
wh
ich
v
alid
ate
o
n
m
u
lti
-
ce
n
ter
E
C
G
d
atasets
,
u
s
in
g
tr
ip
let
lo
s
s
to
s
h
o
w
b
etter
r
esu
lts
o
f
g
en
er
aliza
tio
n
with
d
y
n
am
ic
m
ar
g
in
s
co
n
d
itio
n
ed
o
n
s
tr
ess
s
ev
er
ity
.
W
h
ile
r
ea
l
-
tim
e
E
C
G
m
o
n
ito
r
in
g
in
tr
o
d
u
ce
s
in
wea
r
ab
le
tech
n
o
lo
g
y
s
u
p
p
o
r
ts
n
o
is
e
an
d
m
o
tio
n
d
is
to
r
tio
n
s
w
h
ich
lead
s
t
o
s
m
all
m
o
r
p
h
o
lo
g
ical
tr
aits
[
9
]
‒
[
1
1
]
.
E
v
en
m
o
r
p
h
o
lo
g
ical
ab
b
r
ev
iatio
n
s
ar
e
m
a
k
in
g
m
o
r
e
co
m
p
lex
in
s
ig
n
a
l
in
ter
p
r
etatio
n
b
y
s
tr
ess
f
u
l
cir
cu
m
s
tan
ce
s
,
an
d
th
ese
ar
tifa
cts
ar
e
esp
ec
ially
h
ar
m
f
u
l.
Stro
n
g
b
aselin
e
ac
cu
r
ac
y
i
n
d
etec
tin
g
ar
r
h
y
th
m
ias
h
as
b
ee
n
attain
ed
b
y
co
n
v
en
tio
n
al
C
NN
-
an
d
R
NN
-
b
ased
E
C
G
clas
s
if
ier
s
[
1
2
]
,
[
1
3
]
.
Ho
wev
er
,
th
ese
ar
ch
itectu
r
es
ar
e
s
u
s
ce
p
tib
le
to
m
o
tio
n
a
r
tifa
cts
an
d
s
tr
ess
-
in
d
u
ce
d
v
ar
iab
ilit
y
d
u
e
t
o
th
eir
r
elian
ce
o
n
s
eq
u
en
tial
m
o
d
elin
g
o
r
f
ix
ed
r
ec
ep
tiv
e
f
ield
s
[
1
4
]
.
Ad
d
itio
n
ally
,
th
eir
clin
ical
d
ep
e
n
d
ab
ili
ty
is
lim
ited
d
u
e
to
th
eir
p
o
o
r
in
ter
p
r
etab
ilit
y
[
1
5
]
.
T
h
ese
d
ef
icien
cies
ar
e
ad
d
r
ess
ed
b
y
n
ew
lin
es
o
f
in
q
u
ir
y
.
R
elatio
n
al
an
d
tem
p
o
r
al
r
elatio
n
s
h
ip
s
in
E
C
Gs
ca
n
b
e
m
o
d
eled
b
y
G
NNs,
wh
ich
h
av
e
p
r
o
v
en
to
b
e
m
o
r
e
r
o
b
u
s
t
th
an
tr
a
d
itio
n
al
m
o
d
els
[
5
]
,
[
1
6
]
.
C
o
m
p
ar
ativ
e
an
d
s
elf
-
s
u
p
e
r
v
is
ed
r
ep
r
esen
tatio
n
lear
n
in
g
m
eth
o
d
s
,
s
u
ch
as
p
h
y
s
io
lo
g
ical
tim
e
-
s
er
ies
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
2D
-
CNN
-
GA
C
L
-
E
C
GN
et
g
r
a
p
h
a
tten
tio
n
:
a
r
o
b
u
s
t fra
mewo
r
k
fo
r
elec
tr
o
ca
r
d
io
g
r
a
m
…
(
P.
K
a
vith
a
)
1531
em
b
ed
d
in
g
m
o
d
els
an
d
C
L
OC
S,
h
av
e
also
s
h
o
wn
p
r
o
m
is
e
in
lear
n
in
g
tr
an
s
f
e
r
ab
le
f
ea
t
u
r
es
ac
r
o
s
s
co
n
d
itio
n
s
.
E
C
GNe
t
an
d
m
u
lti
-
s
eg
m
en
t
wav
ef
o
r
m
(
MSW
)
-
tr
an
s
f
o
r
m
e
r
ar
e
s
o
m
e
tr
a
n
s
f
o
r
m
e
r
-
b
ased
E
C
G
s
o
lu
tio
n
s
th
at
ca
n
ad
ap
t
to
v
a
r
y
in
g
s
ig
n
a
l
d
y
n
am
ics
an
d
ex
t
r
ac
t
lo
n
g
-
r
an
g
e
d
e
p
en
d
e
n
cies.
T
h
eir
co
m
b
in
atio
n
f
o
r
s
tr
ess
-
awa
r
e
E
C
G
cla
s
s
if
icatio
n
is
n
o
t y
et
well
ex
p
lo
r
ed
,
h
o
wev
er
.
C
o
n
s
eq
u
en
tly
,
we
r
ed
e
f
in
e
s
tr
ess
d
etec
tio
n
as
an
is
s
u
e
o
f
r
ep
r
esen
tatio
n
lear
n
in
g
f
o
r
p
h
y
s
io
lo
g
ical
d
ata
in
ar
tific
ial
in
tellig
en
ce
(
AI
)
,
wh
er
e
m
o
d
els
s
h
o
u
ld
b
e
r
o
b
u
s
t,
in
ter
p
r
etab
le
,
an
d
g
e
n
er
aliza
b
le
[
1
7
]
.
I
n
o
r
d
er
to
d
o
th
is
,
we
p
r
o
v
id
e
GA
C
L
-
E
C
GNe
t,
wh
ich
co
m
b
in
es
s
tr
ess
-
co
n
d
itio
n
e
d
AC
L
[
1
8
]
,
m
o
r
p
h
o
-
tem
p
o
r
al
g
r
a
p
h
co
n
s
tr
u
ctio
n
,
C
NN
-
b
a
s
ed
d
en
o
is
in
g
,
a
n
d
GAT
-
b
ased
in
ter
p
r
etab
ilit
y
.
New
d
ev
elo
p
m
en
ts
,
esp
ec
iall
y
tr
an
s
f
o
r
m
e
r
ar
ch
itectu
r
es
a
n
d
s
elf
-
s
u
p
er
v
is
ed
lear
n
in
g
m
eth
o
d
s
,
h
av
e
s
h
o
w
n
g
r
ea
t
p
r
o
m
is
e
in
id
en
tify
in
g
lo
n
g
-
r
an
g
e
tem
p
o
r
al
d
e
p
en
d
e
n
cies
an
d
r
ed
u
cin
g
th
e
n
ee
d
f
o
r
lar
g
e
a
n
n
o
tated
d
atasets
[
1
9
]
.
Fed
er
ated
lear
n
in
g
s
y
s
tem
s
ar
e
also
in
cr
ea
s
in
g
ly
s
ig
n
if
ican
t
f
o
r
m
u
lti
-
ce
n
ter
E
C
G
an
aly
s
is
th
a
t
m
ain
tain
s
co
n
f
id
e
n
tial
p
er
s
o
n
al
d
ata
[
2
0
]
.
T
h
ese
p
ar
a
d
ig
m
ch
an
g
es
h
i
g
h
lig
h
t
th
e
s
ig
n
if
ican
ce
an
d
n
o
v
elty
o
f
GACL
-
E
C
GNe
t a
t th
e
s
am
e
tim
e
th
ey
r
esh
ap
e
th
e
f
ield
o
f
E
C
G
in
ter
p
r
etatio
n
.
A
co
n
tex
tu
al
d
escr
ip
tio
n
in
co
r
p
o
r
atin
g
th
ese
tr
en
d
s
is
n
ec
ess
ar
y
to
d
em
o
n
s
tr
ate
h
o
w
GAC
L
-
E
C
GNe
t
b
r
id
g
es
cu
r
r
en
t
g
ap
s
an
d
a
d
v
an
ce
s
th
e
f
ield
.
T
h
e
r
em
ain
d
er
o
f
th
is
p
ap
e
r
is
o
r
g
an
ize
d
as
f
o
llo
ws.
Sectio
n
1
d
is
cu
s
s
es
p
r
o
b
lem
s
s
u
ch
as
n
o
is
e
an
d
in
ter
-
b
ea
t
f
lu
ctu
atio
n
.
Sectio
n
2
p
r
esen
ts
im
p
r
o
v
ed
E
C
G
class
if
icatio
n
u
s
in
g
g
r
ap
h
m
o
d
eli
n
g
an
d
c
o
n
tr
asti
v
e
lear
n
in
g
.
Sectio
n
3
d
escr
ib
es
th
e
p
r
o
p
o
s
ed
2
D
-
C
NN
-
GACL
-
E
C
GNe
t.
Sectio
n
4
p
r
esen
ts
th
e
k
ey
r
esu
lts
an
d
o
b
s
er
v
atio
n
s
.
Sectio
n
5
co
n
cl
u
d
es th
e
s
tu
d
y
.
2.
RE
L
AT
E
D
WO
RK
S
C
NN
an
d
R
NN
ar
ch
itectu
r
es
h
av
e
b
ee
n
wid
ely
u
s
ed
f
o
r
ar
r
h
y
th
m
ia
id
en
tific
atio
n
[
2
1
]
‒
[
2
3
]
.
B
u
t
s
in
ce
th
ey
r
ely
o
n
s
eq
u
en
tial
p
atter
n
s
o
r
f
ix
e
d
r
ec
e
p
tiv
e
f
ield
s
,
th
ey
ar
e
v
u
ln
e
r
ab
le
to
m
o
tio
n
n
o
is
e
an
d
s
tr
ess
-
in
d
u
ce
d
d
is
to
r
tio
n
s
[
2
4
]
,
[
2
5
]
.
GNNs
f
o
r
E
C
G
t
em
p
o
r
al
d
ep
en
d
en
cies
an
d
r
elatio
n
al
p
atter
n
s
in
E
C
G
d
ata
ar
e
r
ep
r
esen
te
d
b
y
g
r
ap
h
m
o
d
els
i
n
clu
d
in
g
atten
tio
n
-
b
ased
GNNs
[
2
6
]
,
ad
a
p
tiv
e
e
d
g
e
-
weig
h
ted
GNNs
[
2
7
]
,
an
d
E
GC
Net.
Nev
er
t
h
el
ess
,
o
n
ly
a
lim
ited
n
u
m
b
er
em
p
lo
y
a
d
ap
tiv
e
ed
g
e
weig
h
ti
n
g
o
r
e
x
p
licitly
tack
le
s
tr
ess
-
in
d
u
ce
d
v
ar
iab
ilit
y
[
1
4
]
.
C
o
n
tr
asti
v
e
an
d
s
elf
-
s
u
p
er
v
is
ed
lear
n
in
g
:
r
o
b
u
s
t
E
C
G
r
ep
r
esen
tatio
n
s
ar
e
ac
q
u
ir
ed
b
y
co
n
tr
asti
v
e
f
r
am
ewo
r
k
s
s
u
ch
as
p
h
y
s
io
lo
g
ical
tim
e
-
s
er
ies
co
n
tr
asti
v
e
lear
n
in
g
an
d
C
L
OC
S.
No
n
eth
eless
,
th
ey
h
av
e
n
o
t
y
et
b
ee
n
a
u
g
m
e
n
ted
to
u
s
e
m
o
r
p
h
o
-
tem
p
o
r
al
g
r
ap
h
s
in
s
tr
ess
-
co
n
d
itio
n
e
d
lear
n
in
g
.
E
C
GNe
t
-
tr
an
s
f
o
r
m
er
s
an
d
MSW
-
t
r
an
s
f
o
r
m
er
a
r
e
n
ew
m
o
d
els
th
at
s
h
o
w
h
o
w
s
elf
-
atten
tio
n
ca
n
f
i
n
d
lo
n
g
-
r
a
n
g
e
d
ep
e
n
d
en
cies.
GACL
-
E
C
GNe
t
b
u
ild
s
o
n
t
h
is
b
y
in
teg
r
atin
g
C
NN
p
r
ep
r
o
ce
s
s
in
g
[
1
0
]
with
GAT
[
2
8
]
a
n
d
AC
L
[
6
]
,
[
1
6
]
.
Fed
er
ated
an
d
eth
ical
AI
in
E
C
G
f
ed
er
ated
lear
n
i
n
g
h
as b
ec
o
m
e
a
well
-
k
n
o
wn
way
t
o
p
r
o
tect
p
r
iv
ac
y
in
E
C
G
[
2
9
]
,
[
3
0
]
.
T
h
e
r
e
ar
e
s
till
p
r
o
b
lem
s
th
at
n
ee
d
to
b
e
wo
r
k
ed
o
u
t
wh
en
it
co
m
es
to
ex
p
lain
ab
ilit
y
,
d
em
o
g
r
ap
h
ic
b
alan
ce
,
an
d
ju
s
tice
[
2
2
]
.
GACL
-
E
C
GNe
t
is
u
s
ef
u
l
in
th
ese
ar
ea
s
b
ec
au
s
e
it
is
f
lex
ib
le
an
d
ca
n
b
e
u
n
d
er
s
to
o
d
.
T
ab
le
1
s
u
m
m
a
r
izes
p
r
ev
io
u
s
E
C
G
class
if
icat
io
n
m
o
d
els
an
d
th
eir
r
elev
an
ce
to
GACL
-
E
C
GNe
t
.
E
v
en
th
o
u
g
h
p
ast
E
C
G
class
if
icatio
n
m
o
d
els
h
av
e
laid
a
s
tr
o
n
g
f
o
u
n
d
a
tio
n
f
o
r
au
to
m
ated
ca
r
d
iac
d
iag
n
o
s
is
,
th
er
e
is
n
o
w
n
o
co
h
esiv
e
n
ar
r
ativ
e
lin
k
i
n
g
th
ese
m
eth
o
d
o
l
o
g
ies
to
th
e
s
p
ec
if
ic
ch
allen
g
es
th
at
GACL
-
E
C
GNe
t
s
ee
k
s
to
ad
d
r
ess
.
T
ab
le
1
.
Su
m
m
a
r
y
o
f
r
elate
d
wo
r
k
s
Li
t
e
r
a
t
u
r
e
A
d
v
a
n
t
a
g
e
s
D
i
sad
v
a
n
t
a
g
e
s
Le
e
e
t
a
l
.
[
5
]
U
ses
a
n
o
v
e
l
g
r
a
p
h
-
b
a
se
d
E
C
G
r
e
p
r
e
sen
t
a
t
i
o
n
w
i
t
h
Q
R
S
-
c
e
n
t
e
r
e
d
p
o
o
l
i
n
g
,
a
c
h
i
e
v
i
n
g
h
i
g
h
a
c
c
u
r
a
c
y
(
M
a
c
r
o
F
1
-
sco
r
e
:
8
8
.
6
1
%)
a
n
d
s
c
a
l
a
b
i
l
i
t
y
o
n
v
a
r
i
a
b
l
e
-
l
e
n
g
t
h
si
g
n
a
l
s.
R
e
l
i
e
s
o
n
p
r
e
c
i
s
e
ma
n
u
a
l
d
e
t
e
c
t
i
o
n
o
f
P
-
Q
R
S
-
T
b
o
u
n
d
a
r
i
e
s,
w
h
i
c
h
l
i
m
i
t
s
a
u
t
o
ma
t
i
o
n
a
n
d
c
o
m
p
l
i
c
a
t
e
s
r
e
a
l
-
t
i
me
c
l
i
n
i
c
a
l
d
e
p
l
o
y
me
n
t
.
M
a
l
l
e
sw
a
r
i
e
t
a
l
.
[
3
1
]
Ef
f
e
c
t
i
v
e
l
y
i
n
t
e
g
r
a
t
e
s
c
o
n
t
i
n
u
o
u
s
w
a
v
e
l
e
t
t
r
a
n
sf
o
r
m
(
C
W
T
)
w
i
t
h
p
r
e
-
t
r
a
i
n
e
d
C
N
N
s
(
S
q
u
e
e
z
e
-
N
e
t
)
a
c
h
i
e
v
i
n
g
h
i
g
h
c
l
a
ssi
f
i
c
a
t
i
o
n
a
c
c
u
r
a
c
y
(
u
p
t
o
9
8
.
7
%)
o
n
EC
G
si
g
n
a
l
s
.
C
o
m
p
u
t
a
t
i
o
n
a
l
o
v
e
r
h
e
a
d
o
f
C
W
T
t
r
a
n
sf
o
r
ma
t
i
o
n
a
n
d
d
e
e
p
m
o
d
e
l
s
l
i
m
i
t
s
r
e
a
l
-
t
i
m
e
d
e
p
l
o
y
me
n
t
a
n
d
g
e
n
e
r
a
l
i
z
a
t
i
o
n
a
c
r
o
ss
su
b
j
e
c
t
s.
Ze
i
n
a
l
i
p
o
u
r
a
n
d
G
o
r
i
[3
2
]
A
c
h
i
e
v
e
s
h
i
g
h
a
c
c
u
r
a
c
y
i
n
E
C
G
u
si
n
g
i
n
n
o
v
a
t
i
v
e
v
i
s
i
b
i
l
i
t
y
g
r
a
p
h
me
t
h
o
d
s
(
n
a
t
u
r
a
l
v
i
si
b
i
l
i
t
y
g
r
a
p
h
(NVG)
,
h
o
r
i
z
o
n
t
a
l
v
i
si
b
i
l
i
t
y
g
r
a
p
h
(
H
V
G
)
,
q
u
a
n
t
i
l
e
g
r
a
p
h
(
Q
G
)
)
c
o
m
b
i
n
e
d
w
i
t
h
g
r
a
p
h
i
so
m
o
r
p
h
i
sm
n
e
t
w
o
r
k
(
G
I
N
)
,
w
i
t
h
o
u
t
r
e
l
y
i
n
g
o
n
man
u
a
l
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
.
La
c
k
s
d
e
n
o
i
s
i
n
g
,
l
e
a
r
n
e
d
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
(
C
N
N
s)
,
a
n
d
s
t
r
e
ss
-
a
w
a
r
e
n
e
ss,
l
i
m
i
t
i
n
g
r
o
b
u
st
n
e
ss
t
o
n
o
i
se
a
n
d
g
e
n
e
r
a
l
i
z
a
t
i
o
n
a
c
r
o
ss
p
h
y
s
i
o
l
o
g
i
c
a
l
c
o
n
d
i
t
i
o
n
s
.
D
e
g
i
r
m
e
n
c
i
e
t
a
l
.
[
3
3
]
A
c
h
i
e
v
e
s
h
i
g
h
a
c
c
u
r
a
c
y
(
9
9
.
7
%)
i
n
a
r
r
h
y
t
h
mi
a
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
s
i
n
g
a
l
i
g
h
t
w
e
i
g
h
t
2D
-
C
N
N
w
i
t
h
mi
n
i
ma
l
p
r
e
p
r
o
c
e
ss
i
n
g
a
n
d
l
o
w
c
o
mp
u
t
a
t
i
o
n
a
l
c
o
s
t
.
R
e
l
i
e
s
o
n
c
o
n
v
e
r
t
i
n
g
1
D
E
C
G
si
g
n
a
l
s t
o
2
D
i
ma
g
e
s
,
w
h
i
c
h
m
a
y
n
o
t
c
a
p
t
u
r
e
t
e
mp
o
r
a
l
f
e
a
t
u
r
e
s
a
s
e
f
f
e
c
t
i
v
e
l
y
a
s
se
q
u
e
n
t
i
a
l
m
o
d
e
l
s
l
i
k
e
R
N
N
s
o
r
l
o
n
g
s
h
o
r
t
-
t
e
r
m
me
mo
r
y
(
LSTM
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
5
2
9
-
1
5
3
8
1532
3.
M
E
T
H
O
D
GACL
-
E
C
GNe
t
is
a
f
u
s
io
n
b
ased
d
ee
p
lear
n
in
g
f
r
a
m
ewo
r
k
s
in
class
if
y
in
g
E
C
Gs
wh
ich
ar
e
d
r
iv
en
b
y
s
tr
ess
.
I
t
f
ir
s
t
co
n
v
er
ts
r
aw
E
C
G
b
ea
ts
in
to
2
D
s
ca
l
o
g
r
am
s
u
s
in
g
C
E
E
MD
AN
in
o
r
d
e
r
to
p
r
eser
v
e
m
o
r
p
h
o
-
tem
p
o
r
al
in
f
o
r
m
atio
n
.
T
h
ese
s
ca
lo
g
r
am
s
ar
e
en
h
an
ce
d
b
y
a
2D
-
C
NN
-
b
ased
d
en
o
is
in
g
m
o
d
u
le
th
at
au
to
m
atica
lly
lear
n
s
n
o
is
e
p
atter
n
s
.
T
h
is
m
eth
o
d
r
em
o
v
es
n
o
is
e
an
d
ar
tifa
cts
f
r
o
m
s
ig
n
al
s
m
o
r
e
e
f
f
ec
tiv
el
y
th
an
s
tan
d
ar
d
f
ilter
s
as
s
h
o
wn
in
Fig
u
r
e
2
.
T
h
e
p
r
o
p
o
s
ed
2D
-
C
NN
-
GACL
-
E
C
G
Net
in
tr
o
d
u
ce
s
s
ev
er
al
m
eth
o
d
o
l
o
g
ical
ad
v
an
ce
m
en
t
s
th
at
p
r
o
p
el
AI
r
esear
ch
f
o
r
war
d
.
W
e
p
r
o
p
o
s
e
a
C
E
E
MD
AN
d
en
o
is
in
g
an
d
m
o
r
p
h
o
-
tem
p
o
r
al
g
r
ap
h
f
o
r
m
u
latio
n
f
o
r
p
h
y
s
io
lo
g
ical
s
ig
n
als,
wh
er
ein
E
C
G
b
ea
ts
ar
e
r
ep
r
esen
ted
as
n
o
d
es
with
h
y
b
r
id
e
d
g
e
weig
h
ts
,
s
p
e
cif
ically
7
0
%
m
o
r
p
h
o
lo
g
ical
s
im
ilar
ity
an
d
3
0
%
tem
p
o
r
al
p
r
o
x
im
ity
.
Yo
u
ca
n
g
et
b
ea
t
-
lev
el
em
b
e
d
d
in
g
s
b
y
u
s
in
g
a
C
NN
-
b
ased
f
ea
tu
r
e
ex
tr
ac
to
r
o
n
d
en
o
is
ed
s
c
alo
g
r
am
s
.
T
o
f
o
r
m
m
o
r
p
h
o
-
tem
p
o
r
al
g
r
ap
h
wh
e
r
e
in
d
iv
id
u
al
b
ea
ts
in
n
o
d
es
a
n
d
e
d
g
es
ar
e
in
p
r
o
p
o
s
ed
m
et
h
o
d
s
ar
e
ap
p
lied
t
o
ex
am
in
e
h
o
w
s
im
ilar
a
n
d
c
o
n
t
in
u
o
u
s
.
W
h
er
ea
s
,
th
e
n
atu
r
al
o
r
d
er
o
f
th
e
s
ig
n
al
is
m
ain
tain
ed
in
s
eq
u
en
tial,
t
o
im
p
r
o
v
es
th
eir
m
o
d
el’
s
ab
ilit
y
to
d
is
tin
g
u
is
h
b
etwe
en
class
es
in
co
n
tr
asti
v
e
lear
n
in
g
ad
ap
ted
ar
e
f
ac
in
g
ch
allen
g
in
g
in
p
air
s
o
f
p
o
s
itiv
e
an
d
n
e
g
ativ
e.
L
im
itatio
n
s
f
o
u
n
d
in
ea
r
lier
m
o
d
els,
a
r
e
d
ir
e
ctly
ad
d
r
ess
es
wea
k
f
ea
tu
r
e
s
ep
ar
atio
n
,
v
u
ln
er
ab
ili
ty
to
n
o
is
e,
in
ad
eq
u
ate
r
elati
o
n
al
m
ap
p
in
g
an
d
lo
s
s
o
f
tem
p
o
r
al
f
id
elity
.
T
h
e
s
y
s
tem
s
ig
n
if
ican
tly
en
h
an
ce
s
s
tr
ess
cla
s
s
if
icatio
n
p
er
f
o
r
m
a
n
ce
in
in
teg
r
atin
g
co
m
p
o
n
e
n
ts
an
d
in
tr
o
d
u
ce
s
a
d
is
tin
ctiv
e
g
r
ap
h
-
b
ased
f
r
a
m
e
wo
r
k
f
o
r
lear
n
in
g
s
tr
ess
-
awa
r
e
r
ep
r
esen
tatio
n
s
o
f
p
h
y
s
io
lo
g
i
ca
l sig
n
als.
Atten
tio
n
-
b
ased
tr
an
s
f
o
r
m
er
m
o
d
els,
alo
n
g
with
o
th
er
r
ec
en
t
im
p
r
o
v
em
e
n
ts
in
in
ter
p
r
e
tab
le
E
C
G
ca
teg
o
r
izatio
n
,
h
a
v
e
s
h
o
wn
p
r
o
m
is
e
in
in
cr
ea
s
in
g
clin
ician
tr
u
s
t.
Fed
er
ated
lear
n
in
g
also
h
elp
s
with
wea
r
ab
le
an
d
p
r
i
v
ac
y
-
s
en
s
itiv
e
E
C
G
an
aly
s
is
b
y
lettin
g
in
s
titu
tio
n
s
tr
ain
s
af
ely
an
d
with
o
u
t
b
ei
n
g
co
n
n
ec
ted
t
o
ea
ch
o
th
er
.
T
h
ese
p
e
r
s
p
ec
tiv
es
wo
u
ld
au
g
m
e
n
t
th
e
an
al
y
tical
f
r
am
ewo
r
k
an
d
s
itu
ate
GACL
-
E
C
GNe
t
with
in
a
b
r
o
ad
e
r
ec
o
s
y
s
tem
o
f
e
v
o
lv
i
n
g
s
o
lu
tio
n
s
.
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
s
u
r
p
ass
es
ex
is
tin
g
g
r
ap
h
-
b
ased
E
C
G
m
o
d
els
b
y
in
c
o
r
p
o
r
atin
g
m
o
r
p
h
o
-
tem
p
o
r
al
e
d
g
e
m
o
d
elin
g
a
n
d
s
tr
ess
-
co
n
d
itio
n
e
d
m
u
lti
-
h
e
ad
atten
tio
n
.
T
h
ese
im
p
r
o
v
em
e
n
ts
m
ak
e
GACL
-
E
C
GNe
t
d
if
f
er
en
t
f
r
o
m
o
ld
er
g
r
a
p
h
c
o
n
v
o
lu
tio
n
al
n
et
wo
r
k
(
GC
N)
-
b
ased
ap
p
r
o
ac
h
es,
wh
ich
o
f
ten
d
i
d
n
'
t
wo
r
k
well
in
clin
ical
s
ettin
g
s
o
r
a
d
ju
s
t
to
d
if
f
er
en
t
tim
e
p
er
io
d
s
.
GACL
-
E
C
GNe
t c
an
r
ea
d
d
ata
in
g
r
ea
t d
etail
an
d
class
if
y
it a
cc
u
r
ately
ev
en
wh
en
th
e
r
e
is
a
lo
t o
f
s
tr
ess
.
Fig
u
r
e
2
.
C
E
E
MD
AN
d
en
o
is
i
n
g
E
C
G
s
ig
n
al
3
.
1
.
2
D
-
CNN
wit
h CEE
M
D
AN
deno
is
i
ng
a
nd
SM
O
T
E
Ad
ap
tiv
e
d
ec
o
m
p
o
s
itio
n
u
s
es
C
E
E
MD
AN
to
s
p
lit
E
C
G
s
ig
n
als
in
to
8
in
tr
i
n
s
ic
m
o
d
e
f
u
n
ctio
n
(
I
MFs
)
,
is
o
latin
g
n
o
is
e
(
h
i
g
h
-
f
r
eq
u
e
n
cy
)
a
n
d
p
h
y
s
io
lo
g
ical
co
m
p
o
n
en
ts
(
lo
w
-
f
r
e
q
u
en
c
y
)
.
I
n
Fig
u
r
e
3
n
o
is
e
r
em
o
v
al
d
is
ca
r
d
s
o
r
th
r
esh
o
l
d
s
I
MFs
1
–
3
,
r
etain
s
I
MFs
4
–
8
,
ac
h
iev
in
g
2
2
%
s
ig
n
al
-
to
-
n
o
is
e
r
atio
(
SNR
)
im
p
r
o
v
em
e
n
t
wh
ile
p
r
eser
v
in
g
s
ig
n
al
f
id
elity
.
Su
p
er
io
r
ity
o
v
er
wav
elets:
o
u
tp
er
f
o
r
m
s
w
av
elet
d
en
o
is
in
g
b
y
8
%
in
s
u
p
p
r
ess
in
g
m
o
ti
o
n
a
r
tifa
cts,
lev
er
ag
in
g
n
o
n
-
lin
ea
r
,
d
ata
-
d
r
iv
en
b
asis
f
u
n
ctio
n
s
.
R
ec
o
n
s
tr
u
cts
clea
n
E
C
G
f
r
o
m
I
MFs
4
–
8
,
m
ai
n
tain
in
g
m
o
r
p
h
o
lo
g
ical
f
ea
tu
r
es
o
f
a
r
r
h
y
t
h
m
ias
an
d
co
n
d
u
ctio
n
b
lo
ck
s
.
I
n
s
tr
ess
-
awa
r
e
cla
s
s
if
icatio
n
b
y
p
r
eser
v
in
g
s
u
b
tle
m
ar
k
er
s
li
k
e
is
ch
em
ic
ST
s
h
if
ts
in
wea
r
ab
le
E
C
G
s
ig
n
al
s
[
2
0
]
,
[
2
1
]
h
as im
p
r
o
v
ed
its
clin
ical
u
tili
ty
.
(
)
=
∑
(
)
,
↑
22%
8
=
1
(
1
)
T
h
e
m
eth
o
d
SMOT
E
is
u
s
ed
t
o
d
ea
l
with
class
im
b
alan
ce
i
n
E
C
G
d
atasets
b
y
p
r
o
d
u
cin
g
s
y
n
th
etic
in
s
tan
ce
s
f
o
r
u
n
d
er
r
e
p
r
esen
ted
class
es su
ch
as v
en
tr
ic
u
lar
ec
to
p
ic
b
ea
t
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
2D
-
CNN
-
GA
C
L
-
E
C
GN
et
g
r
a
p
h
a
tten
tio
n
:
a
r
o
b
u
s
t fra
mewo
r
k
fo
r
elec
tr
o
ca
r
d
io
g
r
a
m
…
(
P.
K
a
vith
a
)
1533
Fig
u
r
e
3
.
Vis
u
aliza
tio
n
o
f
d
en
o
is
in
g
E
C
G
s
ig
n
al
3
.
2
.
G
ra
ph
a
t
t
ent
io
n net
wo
r
k
I
n
GAT
h
ea
r
tb
ea
ts
ar
e
an
aly
ze
d
u
s
in
g
E
C
G
wh
ich
is
s
am
p
le
d
at
3
6
0
Hz,
wh
ic
h
in
tu
r
n
ca
p
tu
r
es
ea
c
h
h
ea
r
tb
ea
t
as
a
2
5
6
s
am
p
les
s
eg
m
en
t.
C
o
n
s
eq
u
en
tly
,
ea
ch
b
ea
t
s
eg
m
en
t
last
s
ap
p
r
o
x
im
a
tely
7
1
1
m
s
w
h
ich
alig
n
s
with
n
o
r
m
al
ca
r
d
iac
f
r
eq
u
en
cy
o
f
6
0
t
o
1
0
0
b
ea
ts
p
er
6
0
s
ec
o
n
d
s
[
2
2
]
.
Usu
ally
th
is
s
am
p
lin
g
r
ate
(
3
6
0
Hz)
is
s
u
f
f
icien
t
to
p
r
o
v
i
d
e
tem
p
o
r
al
r
eso
lu
tio
n
an
d
ca
p
tu
r
in
g
f
in
e
d
etails
s
u
ch
as
P
-
wav
e
m
o
r
p
h
o
lo
g
y
an
d
QR
S
co
m
p
lex
d
u
r
ati
o
n
wh
ich
is
ess
en
tial
f
o
r
class
if
y
in
g
ar
r
h
y
t
h
m
ias
u
n
d
e
r
s
tr
ess
.
E
C
G
b
ea
ts
ar
e
th
e
ed
g
e
weig
h
tin
g
s
ch
em
e
—
7
0
%
m
o
r
p
h
o
lo
g
ical
s
im
ilar
ity
a
n
d
3
0
%
tem
p
o
r
al
p
r
o
x
im
ity
—
was
f
o
u
n
d
b
y
u
s
in
g
g
r
id
s
ea
r
ch
o
p
tim
izatio
n
o
n
th
e
MI
T
-
B
I
H
d
ataset.
T
h
e
s
tu
d
y
ac
r
o
s
s
r
atio
s
o
f
Sen
s
itiv
ity
d
ata
(
e.
g
.
,
5
0
:5
0
an
d
8
0
:2
0
)
h
a
d
th
e
h
ig
h
est
class
if
icatio
n
ac
cu
r
ac
y
in
7
0
:
3
0
E
C
G
d
ata
an
d
th
e
n
s
h
o
wed
th
e
d
ec
r
ea
s
e
r
esu
lt
o
f
f
alse
p
o
s
itiv
es
in
d
ata
wh
en
it
ca
m
e
to
f
in
d
in
g
s
tr
ess
-
in
d
u
ce
d
ar
r
h
y
th
m
ias.
T
h
is
b
alan
ce
d
o
es
a
g
o
o
d
j
o
b
o
f
ca
p
tu
r
in
g
b
o
th
th
e
s
h
ap
e
o
f
t
h
e
wav
ef
o
r
m
an
d
th
e
tim
in
g
b
etwe
en
b
ea
ts
,
wh
ich
ar
e
b
o
t
h
im
p
o
r
tan
t
wh
en
s
tr
ess
lev
els ch
an
g
e.
W
e
em
p
lo
y
ed
o
n
e
-
h
o
t
v
ec
to
r
s
to
en
co
d
e
th
e
s
tr
ess
lab
els,
with
ea
ch
v
ec
to
r
r
ep
r
esen
tin
g
a
d
if
f
er
e
n
t
lev
el
o
f
s
tr
ess
,
s
u
ch
as b
aselin
e,
em
o
tio
n
al,
o
r
co
g
n
itiv
e
.
∈
256
(
2
)
S
im
ilar
ity
an
d
tem
p
o
r
al
ar
e
s
lid
in
g
win
d
o
w
w
=3
m
o
d
els lo
c
al
HR
V
tr
en
d
s
.
C
o
m
b
in
ed
d
is
tan
ce
is
(
3
)
.
(
,
)
=
0
.
7
|
|
,
|
|
2
+
0
.
3
|
−
|
(
3
)
W
h
ich
g
iv
es o
p
tim
ized
v
ia
g
r
i
d
s
ea
r
ch
.
3
.
3
.
2D
-
CNN
-
G
ACL
-
E
CG
Net
T
h
e
ar
ch
itectu
r
e
o
f
2D
-
C
NN
-
GACL
-
E
C
G
NE
T
wh
ich
p
r
o
v
id
es
a
d
etailed
d
escr
ip
tio
n
o
f
m
u
lti
-
h
ea
d
GAT
lay
er
an
d
a
d
ap
tiv
e
c
o
n
tr
asti
v
e
lo
s
s
to
im
p
r
o
v
e
its
tr
an
s
p
ar
en
cy
an
d
r
ep
r
o
d
u
cib
ilit
y
.
I
t
o
u
tlin
es
o
f
ea
ch
lay
er
d
im
en
s
io
n
s
o
f
in
p
u
t
-
o
u
t
p
u
t
ty
p
e,
ac
tiv
atio
n
f
u
n
ctio
n
s
,
an
d
d
r
o
p
o
u
t
co
n
f
ig
u
r
atio
n
s
.
T
h
e
f
u
ll
ar
ch
itectu
r
e
is
s
h
o
wn
in
Fig
u
r
e
4.
3
.
3
.
1
.
M
ulti
-
hea
d
g
ra
ph
a
t
t
e
ntio
n net
wo
rk
l
a
y
er
s
T
h
e
2D
-
C
NN
m
o
d
u
le
co
m
p
r
is
es
th
r
ee
co
n
v
o
lu
tio
n
al
b
l
o
ck
s
(
k
er
n
el
s
ize:
3
×3
,
s
tr
id
e:
1
)
,
ea
ch
f
o
llo
wed
b
y
b
atch
n
o
r
m
aliza
tio
n
an
d
R
eL
U
ac
tiv
atio
n
.
T
h
e
GAT
m
o
d
u
le
in
clu
d
es
f
o
u
r
at
ten
tio
n
h
ea
d
s
with
64
-
d
im
e
n
s
io
n
al
h
id
d
en
lay
e
r
s
,
u
s
in
g
L
ea
k
y
R
eL
U
f
o
r
atten
tio
n
s
co
r
in
g
.
T
h
e
f
in
al
class
if
ic
atio
n
lay
er
em
p
lo
y
s
s
o
f
tm
ax
ac
tiv
atio
n
o
v
er
f
iv
e
b
ea
t
ca
teg
o
r
ies.
Mu
lti
-
h
ea
d
G
AT
lay
er
s
lear
n
n
o
d
e
r
ep
r
ese
n
tatio
n
s
th
at
ar
e
f
u
ll
with
co
n
tex
t.
B
y
p
a
y
in
g
atten
t
io
n
to
s
u
r
r
o
u
n
d
i
n
g
b
ea
ts
th
at
ar
e
r
elev
a
n
t,
th
e
y
ar
e
ab
le
to
s
u
cc
ess
f
u
lly
ca
p
tu
r
e
in
ter
-
b
ea
t d
e
p
en
d
e
n
cy
.
=
(
(
[
ℎ
|
|
ℎ
]
)
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
5
2
9
-
1
5
3
8
1534
3
.
3
.
2
.
Ada
ptiv
e
co
ntr
a
s
t
iv
e
l
o
s
s
Ma
r
g
in
s
lear
n
ed
v
ia
2
-
la
y
er
ML
P (
in
p
u
t: st
r
ess
lab
el,
o
u
tp
u
t: m
ar
g
in
ϵ
[
0
.
2
,
1
.
0
]
)
.
=
0
.
6
+
0
.
4
(
5
)
AC
L
:
o
u
r
s
tr
ess
-
co
n
d
itio
n
ed
AC
L
s
y
s
tem
im
p
r
o
v
es
class
s
ep
ar
ab
ilit
y
b
y
lear
n
in
g
m
ar
g
in
s
b
ased
o
n
th
e
lev
el
o
f
s
tr
ess
.
T
h
ese
tech
n
iq
u
es
ca
n
b
e
u
tili
ze
d
in
m
a
n
y
d
o
m
ain
s
wh
er
e
th
e
s
h
ap
e
an
d
s
ig
n
als
o
f
m
o
r
p
h
o
l
o
g
y
is
af
f
ec
ted
b
y
c
o
n
tex
tu
al
la
b
els,
s
u
ch
as
em
o
tio
n
o
r
wea
r
in
ess
.
I
n
v
ar
io
u
s
d
o
m
ain
s
th
is
ap
p
r
o
ac
h
was
u
s
ed
f
o
r
co
n
tex
tu
al
f
ac
to
r
s
s
u
ch
as
em
o
tio
n
al
s
tate
o
r
f
atig
u
e
wh
ich
ca
n
alter
th
e
m
o
r
p
h
o
lo
g
y
o
f
p
h
y
s
io
lo
g
ical
s
ig
n
als.
B
y
th
e
en
co
d
in
g
s
tr
ateg
y
it
en
h
an
ce
s
,
th
e
d
is
tin
ctio
n
b
et
wee
n
p
ath
o
lo
g
ical
an
d
s
tr
ess
-
r
elate
d
p
atter
n
s
f
o
r
ad
ap
tiv
ely
s
eg
m
en
ti
n
g
E
C
G
b
ea
ts
.
W
e
co
n
d
u
cted
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
s
to
ev
alu
ate
its
ef
f
ec
tiv
en
ess
in
co
m
p
ar
in
g
GACL
-
E
C
GNe
t
ag
ain
s
t
C
NN
-
b
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
B
iLST
M
)
an
d
GC
N
m
o
d
els
o
v
er
th
e
f
o
ld
s
o
f
f
iv
e
c
r
o
s
s
-
v
alid
atio
n
f
o
ld
s
.
Per
f
o
r
m
an
ce
r
esu
lts
y
ield
ed
s
tati
s
tically
s
ig
n
if
ican
t
im
p
r
o
v
em
e
n
t,
with
all
p
-
v
alu
e
s
f
allin
g
b
elo
w
0
.
0
1
.
Ad
d
itio
n
ally
,
we
ev
alu
ated
th
e
m
o
d
el
’
s
ab
ilit
y
to
s
p
ec
if
y
ac
r
o
s
s
d
atasets
b
y
t
r
ain
in
g
o
n
MI
T
-
B
I
H
a
n
d
test
in
g
o
n
we
ar
ab
le
s
tr
ess
an
d
af
f
ec
t
d
etec
t
io
n
(
W
E
SAD)
an
d
s
tr
ess
at
wo
r
k
,
k
n
o
wled
g
e
wo
r
k
(
SW
E
L
L
-
KW
)
,
r
ev
ea
lin
g
r
o
b
u
s
t c
r
o
s
s
-
d
o
m
ain
p
er
f
o
r
m
an
c
e.
3
.
3
.
3
.
MIT
-
B
I
H
a
rr
hy
t
hm
ia
da
t
a
ba
s
e
A
class
im
b
alan
ce
,
m
ar
k
ed
w
ith
m
in
o
r
ity
class
es
s
u
ch
as
APC
co
m
p
r
is
in
g
o
n
ly
0
.
6
%
o
f
all
b
ea
ts
.
C
o
llected
b
aselin
e
(
n
o
n
-
s
tr
ess
)
co
n
d
itio
n
s
,
th
e
d
atasets
s
e
r
v
in
g
as
a
r
ef
er
e
n
ce
f
o
r
d
etec
tio
n
o
f
ar
r
h
y
th
m
i
a
u
n
d
er
m
o
n
ito
r
ed
cir
cu
m
s
tan
ce
s
.
Data
s
ets
wh
ich
co
m
p
r
i
s
e
PVC
,
R
B
B
B
,
L
B
B
B
,
al
o
n
g
with
APC
,
o
f
4
7
in
d
iv
id
u
als
(
2
5
m
ales
an
d
2
2
f
em
ales)
a
r
e
th
e
two
c
o
n
t
r
ac
tio
n
s
th
at
f
all
in
to
th
e
f
iv
e
clin
ically
r
elev
a
n
t
ca
teg
o
r
ies
with
ag
es
r
an
g
in
g
f
r
o
m
2
3
t
o
8
9
y
ea
r
s
th
at
ar
e
an
n
o
tated
in
to
th
e
E
C
G
r
ec
o
r
d
in
g
s
.
All
r
ec
o
r
d
in
g
s
ar
e
s
am
p
led
at
3
6
0
Hz,
with
a
to
tal
o
f
1
0
9
,
4
4
6
b
ea
ts
.
3
.
3
.
4
.
Wea
ra
ble st
re
s
s
a
nd
a
f
f
ec
t
det
ec
t
io
n da
t
a
s
et
Per
f
o
r
m
in
g
s
tr
o
o
p
test
with
E
C
G
r
ec
o
r
d
in
g
s
o
f
2
5
o
f
f
ice
w
o
r
k
er
s
i
n
d
esk
-
b
ased
j
o
b
s
.
T
h
e
c
o
g
n
itiv
e
wo
r
k
lo
ad
was
class
if
ied
in
to
t
h
r
ee
lev
els:
lo
w,
m
ed
iu
m
,
a
n
d
h
ig
h
.
E
C
G
s
ig
n
als
ar
e
r
ec
o
r
d
ed
at
2
,
0
4
8
Hz,
an
d
th
en
later
s
am
p
led
d
o
wn
to
3
6
0
Hz
f
o
r
c
o
n
s
is
ten
cy
ac
r
o
s
s
d
atasets
.
H
R
V
v
ar
iatio
n
s
an
d
ST
-
T
s
eg
m
en
t
s
h
if
ts
p
h
y
s
io
lo
g
ical
i
n
d
icato
r
s
in
s
u
b
tle
ar
e
ass
o
ciate
d
with
co
g
n
itiv
e
lo
ad
s
wer
e
r
ec
o
r
d
ed
to
s
u
p
p
o
r
t
co
n
tex
t
s
tr
ess
s
en
s
itiv
e
E
C
G
ev
alu
atio
n
.
Fig
u
r
e
4
.
GACL
-
E
C
G
Net
ar
ch
itectu
r
e
f
o
r
s
tr
ess
-
awa
r
e
E
C
G
s
ig
n
al
class
if
icatio
n
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
n
T
ab
le
2
f
o
r
s
tr
ess
class
if
icatio
n
an
d
a
r
r
h
y
th
m
ia
d
etec
ti
o
n
ar
e
c
o
m
p
ar
e
d
to
g
et
th
e
ev
alu
atio
n
ag
ain
s
t
s
ev
er
al
b
aselin
es
o
f
GACL
-
E
C
G
Net.
T
o
test
s
tatis
tica
lly
we
allo
wed
to
em
p
lo
y
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
th
e
s
tu
d
y
o
f
f
iv
e
f
o
ld
s
t
o
co
m
p
ar
e
o
u
t
p
er
f
o
r
m
an
ce
o
f
GACL
-
E
C
GNe
t
ag
ain
s
t
C
NN
-
B
iLST
M
an
d
GC
N
b
aselin
es,
wh
ich
s
tatis
tically
s
h
o
wed
th
e
p
e
r
f
o
r
m
an
ce
u
p
g
r
a
d
es.
T
h
is
s
h
o
ws
th
at
all
co
m
p
ar
is
o
n
s
y
ield
u
n
d
er
0
.
0
1
P
v
alu
es
a
n
aly
s
is
s
h
o
win
g
m
ar
k
ed
e
n
h
an
ce
m
en
ts
in
o
v
er
all
p
er
f
o
r
m
an
ce
.
T
h
e
s
tu
d
y
ex
am
in
ed
cr
o
s
s
-
d
ataset
g
en
e
r
aliza
tio
n
b
y
u
s
in
g
MI
T
-
B
I
H
f
o
r
tr
ain
in
g
an
d
e
v
alu
atin
g
p
er
f
o
r
m
an
ce
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
2D
-
CNN
-
GA
C
L
-
E
C
GN
et
g
r
a
p
h
a
tten
tio
n
:
a
r
o
b
u
s
t fra
mewo
r
k
fo
r
elec
tr
o
ca
r
d
io
g
r
a
m
…
(
P.
K
a
vith
a
)
1535
W
E
SAD
an
d
SW
E
L
L
-
K
W
d
atasets
.
T
o
h
an
d
le
b
o
th
p
h
y
s
io
lo
g
ical
an
d
p
s
y
ch
o
l
o
g
ical
s
tr
ess
d
etec
tio
n
task
s
th
e
m
o
d
el
h
as
ab
ilit
y
to
m
ain
tain
ed
ab
o
v
e
9
0
%
in
F1
-
s
co
r
e.
W
e
co
n
d
u
cted
co
m
p
ar
i
s
o
n
test
s
u
til
izin
g
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
(
e
.
g
.
,
E
C
G
-
B
E
R
T
an
d
tim
e
-
s
er
ies
tr
an
s
f
o
r
m
er
)
,
s
elf
-
s
u
p
e
r
v
is
ed
f
r
am
ew
o
r
k
s
(
e.
g
.
,
SimCLR
an
d
Mo
C
o
)
,
a
n
d
h
y
b
r
id
a
r
ch
itectu
r
es
(
e.
g
.
,
C
NN+
L
STM
an
d
GC
N+
t
r
an
s
f
o
r
m
er
)
to
ev
alu
ate
th
e
b
r
o
a
d
er
AI
s
ig
n
if
ican
ce
o
f
GACL
-
E
C
GNe
t.
T
h
e
GACL
-
E
C
GNe
t
m
o
d
el
o
u
tp
er
f
o
r
m
ed
all
b
aselin
e
m
eth
o
d
s
in
s
tr
ess
class
if
icatio
n
task
s
b
y
d
eliv
er
in
g
b
etter
F1
-
s
co
r
es
an
d
im
p
r
o
v
ed
r
esu
lts
.
T
h
e
r
esu
lts
s
h
o
w
th
at
th
e
m
o
d
el
p
e
r
f
o
r
m
s
at
a
co
m
p
etitiv
e
lev
el
f
o
r
b
io
m
e
d
ical
en
g
in
ee
r
in
g
task
s
an
d
m
atc
h
es
th
e
p
er
f
o
r
m
an
ce
o
f
s
tan
d
ar
d
AI
tim
e
-
s
er
ies lea
r
n
in
g
b
e
n
ch
m
ar
k
s
.
T
ab
le
2
.
B
en
ch
m
a
r
k
in
g
o
f
d
if
f
er
en
t E
C
G
d
atasets
M
o
d
e
l
M
I
T
-
B
I
H
(
F
1
-
sco
r
e
)
W
ESA
D
(
F
1
-
sc
o
r
e
)
S
W
ELL
-
K
W
(
F
1
)
C
N
N
-
B
i
LS
TM
9
6
.
2
%
±
1
.
1
8
9
.
1
%
±
2
.
3
8
7
.
3
%
±
1
.
8
G
C
N
9
7
.
1
%
±
0
.
8
9
1
.
4
%
±
1
.
7
8
9
.
4
%
±
1
.
5
2D
-
C
N
N
-
G
A
C
L
-
EC
G
N
e
t
9
8
.
7
%
±
0
.
3
9
4
.
2
%
±
1
.
2
9
2
.
8
%
±
1
.
0
Fig
u
r
e
5
p
r
esen
t
F1
-
s
co
r
e
f
o
r
m
o
d
el
with
d
if
f
er
e
n
t
d
atasets
.
T
h
ese
m
ea
s
u
r
em
en
ts
p
r
o
v
id
e
a
d
ee
p
er
in
s
ig
h
t
o
f
m
o
d
el
b
eh
a
v
io
r
,
es
p
ec
ially
f
o
r
m
i
n
o
r
ity
class
es
lik
e
APC
an
d
L
B
B
B
.
W
e
te
s
t
ed
m
em
o
r
y
u
s
e
an
d
in
f
er
en
ce
laten
cy
o
n
m
o
b
ile
GPUs
(
NVI
DI
A
J
etso
n
Nan
o
,
Ad
r
en
o
6
6
0
)
to
s
ee
if
t
h
ey
c
o
u
ld
b
e
u
s
ed
in
th
e
f
ield
.
T
h
e
m
o
d
el
p
er
f
o
r
m
s
b
e
tter
th
an
th
e
GC
N
an
d
C
NN
-
B
iLST
M
b
aselin
es,
with
a
lat
en
cy
o
f
1
2
m
s
p
er
b
ea
t
an
d
a
m
em
o
r
y
u
s
ag
e
o
f
2
8
MB.
Fig
u
r
e
6
r
esu
lts
s
h
o
w
th
at
GACL
-
E
C
GNe
t
ca
n
b
e
u
s
ed
to
d
etec
t
s
tr
es
s
in
r
ea
l
tim
e
o
n
wea
r
a
b
le
an
d
ed
g
e
p
latf
o
r
m
s
.
T
h
e
p
r
im
ar
y
ch
allen
g
e
in
AI
is
to
lear
n
s
tr
o
n
g
r
e
p
r
esen
tatio
n
s
f
r
o
m
p
h
y
s
io
lo
g
ical
in
p
u
ts
in
e
n
v
ir
o
n
m
en
ts
th
at
ar
e
n
o
is
y
,
d
y
n
am
ic,
an
d
in
f
lu
e
n
ce
d
b
y
s
tr
ess
,
ev
en
if
th
e
m
ain
p
u
r
p
o
s
e
o
f
th
e
s
tu
d
y
is
to
u
s
e
E
C
G
to
f
in
d
s
tr
e
s
s
.
Fig
u
r
e
7
atten
tio
n
m
ap
s
as
a
tech
n
i
q
u
e
o
f
ex
p
lain
a
b
le
ar
tific
ial
in
tellig
en
ce
(
XAI
)
:
t
o
m
ak
e
AI
ea
s
ier
to
u
n
d
er
s
ta
n
d
(
XAI
)
in
h
ea
lth
ca
r
e
d
iag
n
o
s
tics
,
m
u
lti
-
h
ea
d
GAT
lay
er
s
m
ak
e
atten
tio
n
h
ea
tm
ap
s
th
at
a
r
e
ea
s
y
to
u
n
d
e
r
s
tan
d
an
d
m
atch
clin
ical
m
ar
k
er
s
(
s
u
ch
ST
-
s
eg
m
en
t sh
if
ts
)
.
B
ias
an
d
f
air
n
ess
:
p
h
y
s
io
lo
g
i
ca
l
s
tatis
tic
s
o
f
ten
s
h
o
w
d
e
m
o
g
r
ap
h
ic
im
b
alan
ce
s
,
lik
e
d
if
f
er
en
ce
s
in
ag
e,
g
en
d
er
,
a
n
d
h
o
w
p
eo
p
le
r
esp
o
n
d
t
o
s
tr
ess
.
W
e
u
s
e
s
t
r
atif
ied
s
am
p
lin
g
an
d
ex
p
er
t
ev
alu
atio
n
to
c
h
ec
k
s
y
n
th
etic
b
ea
ts
an
d
m
ak
e
s
u
r
e
th
at
all
cla
s
s
if
icat
io
n
s
ar
e
f
air
.
R
ea
l
-
wo
r
ld
d
ep
lo
y
m
en
t:
t
o
m
ak
e
th
e
m
o
d
el
b
etter
f
o
r
e
d
g
e
d
e
p
lo
y
m
en
t,
T
o
r
ch
Scr
ip
t
,
a
n
d
q
u
an
tizatio
n
ar
e
em
p
l
o
y
ed
.
T
h
is
m
ak
es
it
p
o
s
s
ib
le
to
d
o
in
f
er
en
ce
q
u
ick
ly
o
n
m
o
b
ile
GPUs
an
d
R
asp
b
er
r
y
Pi.
T
h
is
m
ak
es
it
ea
s
ier
f
o
r
wea
r
ab
le
eq
u
ip
m
en
t
to
ass
es
s
s
tr
ess
in
r
ea
l
tim
e,
wh
ich
co
u
ld
b
e
u
s
ef
u
l
in
telem
ed
icin
e
a
n
d
m
e
n
tal
h
ea
lth
s
cr
ee
n
in
g
.
F
u
tu
r
e
r
esear
c
h
will
ex
am
in
e
f
ed
er
ated
lear
n
in
g
t
o
p
r
o
v
id
e
s
ec
u
r
e
,
m
u
lti
-
ce
n
te
r
tr
ain
in
g
wh
ile
s
af
eg
u
ar
d
in
g
p
atien
t
p
r
iv
ac
y
,
in
ac
co
r
d
an
ce
with
eth
ical
s
tan
d
ar
d
s
in
AI
f
o
r
h
ea
lth
ca
r
e.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
d
e
m
o
n
s
tr
ates
s
tr
o
n
g
clin
ical
u
tili
ty
b
y
ef
f
ec
tiv
ely
d
etec
tin
g
s
tr
ess
-
ag
g
r
av
ated
a
r
r
h
y
t
h
m
i
as,
s
u
ch
as
Prin
z
m
etal
an
g
in
a,
o
r
s
u
d
d
en
ca
r
d
iac
d
ea
th
.
T
o
v
alid
ate
th
e
in
ter
p
r
etab
ilit
y
an
d
r
eliab
ilit
y
o
f
th
e
m
o
d
el,
we
em
p
lo
y
ed
atten
tio
n
m
ap
v
is
u
aliza
tio
n
tech
n
iq
u
es a
n
d
th
en
v
alid
ated
u
s
in
g
c
lin
ical
d
ata
in
Fig
u
r
e
8
.
Fig
u
r
e
5
.
Mo
d
el
p
er
f
o
r
m
a
n
ce
(
F1
-
s
co
r
es)
ac
r
o
s
s
E
C
G
d
atasets
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
5
2
9
-
1
5
3
8
1536
Fig
u
r
e
6
.
Me
tr
ics f
o
r
co
m
p
u
ta
tio
n
al
ef
f
icien
cy
Fig
u
r
e
7
.
Atten
tio
n
h
ea
tm
ap
o
v
er
E
C
G
s
ig
n
al
Fig
u
r
e
8
.
C
lin
ical
p
er
f
o
r
m
an
c
e
co
m
p
ar
is
o
n
:
p
r
ec
is
io
n
an
d
f
a
ls
e
p
o
s
itiv
es a
cr
o
s
s
m
o
d
els
5.
CO
NCLU
SI
O
N
GACL
-
E
C
GNe
t
en
h
an
ce
s
AI
f
o
r
m
o
d
elin
g
tim
e
-
s
er
ies
an
d
p
h
y
s
io
lo
g
ical
s
ig
n
als
b
y
t
h
e
i
n
teg
r
atio
n
o
f
in
ter
p
r
etab
le
atten
tio
n
m
ec
h
an
is
m
s
,
ad
ap
tiv
e
co
n
t
r
asti
v
e
lo
s
s
,
an
d
g
r
ap
h
-
b
ased
lear
n
in
g
.
B
y
co
m
b
in
in
g
AC
L
with
in
ter
p
r
eta
b
le
GA
T
s
,
o
u
r
m
o
d
el
ac
h
iev
ed
to
p
-
n
o
tch
p
e
r
f
o
r
m
an
ce
o
n
b
o
t
h
ar
r
h
y
th
m
ia
an
d
s
tr
ess
-
in
d
u
ce
d
E
C
G
d
atasets
.
E
ac
h
m
o
d
u
le
s
ig
n
if
ican
tly
en
h
an
ce
d
to
th
e
m
o
d
el'
s
ac
cu
r
ac
y
an
d
in
ter
p
r
etab
ilit
y
in
v
alid
atin
g
ab
latio
n
s
tu
d
ies,
wh
er
e
ea
ch
co
m
p
o
n
e
n
t
is
in
v
er
y
h
ig
h
ac
cu
r
ac
y
,
g
en
er
aliza
b
ilit
y
,
an
d
clin
ical
in
ter
p
r
etab
ilit
y
,
b
ey
o
n
d
th
e
s
tr
ess
d
etec
tio
n
with
wid
e
r
a
n
g
e
o
f
ap
p
licatio
n
lik
e
em
o
tio
n
r
ec
o
g
n
itio
n
,
f
atig
u
e
m
o
n
ito
r
in
g
,
an
d
m
u
ltimo
d
al
h
ea
lth
an
aly
tics
in
s
u
p
p
o
r
tin
g
A
I
f
r
am
ewo
r
k
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
2D
-
CNN
-
GA
C
L
-
E
C
GN
et
g
r
a
p
h
a
tten
tio
n
:
a
r
o
b
u
s
t fra
mewo
r
k
fo
r
elec
tr
o
ca
r
d
io
g
r
a
m
…
(
P.
K
a
vith
a
)
1537
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
wo
u
ld
lik
e
to
ac
k
n
o
wled
g
e
th
e
s
u
p
p
o
r
t
an
d
f
ac
ilit
ies
p
r
o
v
id
ed
b
y
Pre
s
id
en
cy
Un
iv
er
s
ity
f
o
r
th
is
r
esear
ch
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
s
tu
d
y
was c
o
n
d
u
cted
with
o
u
t a
n
y
f
in
an
cial
s
p
o
n
s
o
r
s
h
ip
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
P.
Kav
ith
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
L.
Sh
ak
k
ee
r
a
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
er
e
ar
e
n
o
co
m
p
etin
g
i
n
ter
e
s
ts
to
d
is
clo
s
e
in
th
is
r
esear
ch
.
DATA AV
AI
L
AB
I
L
I
T
Y
No
n
ew
d
ata
wer
e
cr
ea
ted
o
r
a
n
aly
ze
d
in
t
h
is
r
esear
ch
.
RE
F
E
R
E
NC
E
S
[
1
]
R
.
C
h
e
n
g
,
Z
.
Zh
u
a
n
g
,
S
.
Z
h
u
a
n
g
,
L.
X
i
e
,
a
n
d
J.
G
u
o
,
“
M
S
W
-
t
r
a
n
sf
o
r
mer:
mu
l
t
i
-
sca
l
e
s
h
i
f
t
e
d
w
i
n
d
o
w
s
t
r
a
n
sf
o
r
mer
n
e
t
w
o
r
k
s
f
o
r
12
-
l
e
a
d
EC
G
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
2
0
2
3
,
a
r
X
i
v
:
2
3
0
6
.
1
2
0
9
8
.
[
2
]
Y
.
A
mek
r
a
n
,
N
.
D
a
mo
u
n
,
a
n
d
A
.
J.
El
H
a
n
g
o
u
c
h
e
,
“
A
n
a
l
y
si
s
o
f
f
r
e
q
u
e
n
c
y
-
d
o
m
a
i
n
h
e
a
r
t
r
a
t
e
v
a
r
i
a
b
i
l
i
t
y
u
s
i
n
g
a
b
so
l
u
t
e
v
e
r
s
u
s
n
o
r
m
a
l
i
z
e
d
v
a
l
u
e
s
:
i
m
p
l
i
c
a
t
i
o
n
s
a
n
d
p
r
a
c
t
i
c
a
l
c
o
n
c
e
r
n
s,
”
Fro
n
t
i
e
rs
i
n
Ph
y
si
o
l
o
g
y
,
v
o
l
.
1
5
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
8
9
/
f
p
h
y
s
.
2
0
2
4
.
1
4
7
0
6
8
4
.
[
3
]
S
.
S
i
e
c
i
ń
sk
i
,
E.
J
.
T
k
a
c
z
,
a
n
d
P
.
S
.
K
o
st
k
a
,
“
H
e
a
r
t
r
a
t
e
v
a
r
i
a
b
i
l
i
t
y
a
n
a
l
y
s
i
s
o
n
e
l
e
c
t
r
o
c
a
r
d
i
o
g
r
a
ms
,
se
i
sm
o
c
a
r
d
i
o
g
r
a
ms
a
n
d
g
y
r
o
c
a
r
d
i
o
g
r
a
ms
o
f
h
e
a
l
t
h
y
v
o
l
u
n
t
e
e
r
s
a
n
d
p
a
t
i
e
n
t
s
w
i
t
h
v
a
l
v
u
l
a
r
h
e
a
r
t
d
i
se
a
ses
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
3
,
n
o
.
4
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
3
0
4
2
1
5
2
.
[
4
]
W
.
S
e
o
,
N
.
K
i
m,
C
.
P
a
r
k
,
a
n
d
S
.
M
.
P
a
r
k
,
“
D
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
d
e
t
e
c
t
i
n
g
w
o
r
k
-
r
e
l
a
t
e
d
s
t
r
e
ss
u
si
n
g
m
u
l
t
i
mo
d
a
l
si
g
n
a
l
s,”
I
EEE
S
e
n
so
r
s J
o
u
rn
a
l
,
v
o
l
.
2
2
,
n
o
.
1
2
,
p
p
.
1
1
1
2
4
–
1
1
1
3
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
EN
.
2
0
2
2
.
[
5
]
M
.
L
e
e
,
J.
L
i
m,
a
n
d
J
.
K
i
m
,
“
E
C
G
-
G
r
a
p
h
N
e
t
:
a
d
v
a
n
c
e
d
a
r
r
h
y
t
h
mi
a
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
a
s
e
d
o
n
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s,”
H
e
a
r
t
Rh
y
t
h
m
O
2
,
v
o
l
.
6
,
n
o
.
8
,
p
p
.
1
1
9
9
–
1
2
1
1
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
h
r
o
o
.
2
0
2
5
.
0
5
.
0
1
2
.
[
6
]
D
.
K
i
y
a
sse
h
,
T.
Zh
u
,
a
n
d
D
.
A
.
C
l
i
f
t
o
n
,
“
C
LO
C
S
:
c
o
n
t
r
a
s
t
i
v
e
l
e
a
r
n
i
n
g
o
f
c
a
r
d
i
a
c
si
g
n
a
l
s
a
c
r
o
ss
sp
a
c
e
,
t
i
me
,
a
n
d
p
a
t
i
e
n
t
s,”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
Re
sea
r
c
h
,
2
0
2
1
,
p
p
.
5
6
0
6
–
5
6
1
5
.
[
7
]
J.
P
e
n
g
,
A
.
R
a
n
,
C
.
Y
u
,
a
n
d
H
.
Li
u
,
“
EG
C
N
e
t
:
a
h
i
e
r
a
r
c
h
i
c
a
l
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
i
m
p
r
o
v
e
d
c
l
a
ssi
f
i
c
a
t
i
o
n
o
f
e
l
e
c
t
r
o
c
a
r
d
i
o
g
r
a
ms,
”
EU
RA
S
I
P
J
o
u
r
n
a
l
o
n
A
d
v
a
n
c
e
s
i
n
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
,
v
o
l
.
2
0
2
4
,
n
o
.
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s
1
3
6
3
4
-
0
2
4
-
0
1
1
8
7
-
3.
[
8
]
M
.
B
.
A
b
u
b
a
k
e
r
a
n
d
B
.
B
a
b
a
y
i
ğ
i
t
,
“
D
e
t
e
c
t
i
o
n
o
f
c
a
r
d
i
o
v
a
sc
u
l
a
r
d
i
s
e
a
s
e
s
i
n
EC
G
i
ma
g
e
s
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
met
h
o
d
s
,
”
I
EE
E
T
r
a
n
sa
c
t
i
o
n
s
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
,
n
o
.
2
,
p
p
.
3
7
3
–
3
8
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TA
I
.
2
0
2
2
.
3
1
5
9
5
0
5
.
[
9
]
H
.
B
e
c
h
i
n
i
a
,
D
.
B
e
n
merz
o
u
g
,
a
n
d
N
.
K
h
l
i
f
a
,
“
A
p
p
r
o
a
c
h
b
a
se
d
l
i
g
h
t
w
e
i
g
h
t
c
u
s
t
o
m
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
a
n
d
f
i
n
e
-
t
u
n
e
d
M
o
b
i
l
e
N
e
t
-
V
2
f
o
r
E
C
G
a
r
r
h
y
t
h
mi
a
si
g
n
a
l
s
c
l
a
ss
i
f
i
c
a
t
i
o
n
.
,
”
I
EE
E
A
c
c
e
s
s
,
v
o
l
.
1
2
,
p
p
.
4
0
8
2
7
–
4
0
8
4
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
3
7
8
7
3
0
.
[
1
0
]
H
.
Li
n
,
K
.
C
h
e
n
,
Y
.
X
u
e
,
S
.
Zh
o
n
g
,
L.
C
h
e
n
,
a
n
d
M
.
Y
e
,
“
C
o
r
o
n
a
r
y
h
e
a
r
t
d
i
se
a
se
p
r
e
d
i
c
t
i
o
n
met
h
o
d
f
u
si
n
g
d
o
m
a
i
n
-
a
d
a
p
t
i
v
e
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
w
i
t
h
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s
(
G
C
N
)
,
”
S
c
i
e
n
t
i
f
i
c
R
e
p
o
r
t
s
,
v
o
l
.
1
3
,
n
o
.
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
3
8
/
s4
1
5
9
8
-
023
-
3
3
1
2
4
-
z.
[
1
1
]
O
.
C
.
P
h
u
k
a
n
,
A
.
D
a
s,
A
.
B
.
B
u
d
u
r
u
,
a
n
d
R
.
S
h
a
r
ma,
“
S
O
N
I
C
:
sy
n
e
r
g
i
z
i
n
g
v
i
s
i
o
n
f
o
u
n
d
a
t
i
o
n
m
o
d
e
l
s
f
o
r
s
t
r
e
ss
r
e
c
o
g
n
i
t
i
o
n
f
r
o
m
EC
G
s
i
g
n
a
l
s,
”
E
u
r
o
p
e
a
n
S
i
g
n
a
l
P
ro
c
e
ssi
n
g
C
o
n
f
e
r
e
n
c
e
,
p
p
.
1
4
2
1
–
1
4
2
5
,
2
0
2
4
,
d
o
i
:
1
0
.
2
3
9
1
9
/
e
u
si
p
c
o
6
3
1
7
4
.
2
0
2
4
.
1
0
7
1
5
2
6
8
.
[
1
2
]
R
.
S
a
d
i
g
a
l
e
,
S
.
N
a
l
e
,
P
.
K
a
m
b
l
e
,
a
n
d
R
.
W
a
n
a
v
e
,
“
A
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
t
o
EC
G
c
l
a
ssi
f
i
c
a
t
i
o
n
f
o
r
a
t
r
i
a
l
f
i
b
r
i
l
l
a
t
i
o
n
,
n
o
r
m
a
l
si
n
u
s r
h
y
t
h
m,
a
n
d
c
o
n
g
e
s
t
i
v
e
h
e
a
r
t
f
a
i
l
u
r
e
,
”
Re
s
e
a
r
c
h
S
q
u
a
r
e
,
p
p
.
1
–
2
7
,
2
0
2
3
,
d
o
i
:
1
0
.
2
1
2
0
3
/
r
s
.
3
.
r
s
-
3
6
9
3
4
4
9
/
v
1
.
[
1
3
]
M
.
N
a
j
i
a
a
n
d
B
.
F
a
o
u
z
i
,
“
A
n
e
n
h
a
n
c
e
d
h
y
b
r
i
d
m
o
d
e
l
c
o
mb
i
n
i
n
g
C
N
N
,
B
i
LST
M
,
a
n
d
a
t
t
e
n
t
i
o
n
mec
h
a
n
i
sm
f
o
r
EC
G
se
g
me
n
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
Bi
o
m
e
d
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
a
n
d
C
o
m
p
u
t
a
t
i
o
n
a
l
Bi
o
l
o
g
y
,
v
o
l
.
1
6
,
2
0
2
5
,
d
o
i
:
1
0
.
1
1
7
7
/
1
1
7
9
5
9
7
2
2
5
1
3
4
1
0
5
1
.
[
1
4
]
K
.
K
.
C
.
R
e
d
d
y
e
t
a
l
.
,
“
A
mu
l
t
i
-
sca
l
e
c
o
n
v
o
l
u
t
i
o
n
a
l
LST
M
-
d
e
n
s
e
n
e
t
w
o
r
k
f
o
r
r
o
b
u
st
c
a
r
d
i
a
c
a
r
r
h
y
t
h
m
i
a
c
l
a
ssi
f
i
c
a
t
i
o
n
f
r
o
m
EC
G
si
g
n
a
l
s,
”
C
o
m
p
u
t
e
rs
i
n
B
i
o
l
o
g
y
a
n
d
M
e
d
i
c
i
n
e
,
v
o
l
.
1
9
1
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
b
i
o
m
e
d
.
2
0
2
5
.
1
1
0
1
2
1
.
[
1
5
]
S
.
S
e
o
n
i
e
t
a
l
.
,
“
A
p
p
l
i
c
a
t
i
o
n
o
f
s
p
a
t
i
a
l
u
n
c
e
r
t
a
i
n
t
y
p
r
e
d
i
c
t
o
r
i
n
C
N
N
-
B
i
LS
TM
mo
d
e
l
u
si
n
g
c
o
r
o
n
a
r
y
a
r
t
e
r
y
d
i
se
a
se
EC
G
si
g
n
a
l
s
,
”
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
6
6
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
s
.
2
0
2
4
.
1
2
0
3
8
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
5
2
9
-
1
5
3
8
1538
[
1
6
]
A
.
B
.
S
l
a
ma
,
Y
.
A
mr
i
,
A
.
F
n
a
i
e
c
h
,
a
n
d
H
.
S
a
h
l
i
,
“
A
u
t
o
m
a
t
e
d
EC
G
a
r
r
h
y
t
h
m
i
a
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
h
y
b
r
i
d
C
N
N
-
S
V
M
a
r
c
h
i
t
e
c
t
u
r
e
s
,
”
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
o
n
i
c
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
3
,
n
o
.
3
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
n
l
e
st
.
2
0
2
5
.
1
0
0
3
1
6
.
[
1
7
]
S
.
S
a
n
k
a
r
,
M
.
Le
w
i
s,
a
n
d
P
.
H
o
se
i
n
,
“
A
b
o
v
e
g
r
o
u
n
d
b
i
o
mass
e
st
i
ma
t
i
o
n
o
f
a
c
o
c
o
a
p
l
a
n
t
a
t
i
o
n
u
si
n
g
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
i
n
2
0
2
2
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
D
a
t
a
Mi
n
i
n
g
Wo
r
k
s
h
o
p
s (I
C
D
MW
)
,
2
0
2
2
,
p
p
.
1
–
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
D
M
W
5
8
0
2
6
.
2
0
2
2
.
0
0
1
4
7
.
[
1
8
]
J.
Q
i
n
a
n
d
F
.
L
i
u
,
“
G
A
F
-
F
u
si
o
n
N
e
t
:
mu
l
t
i
mo
d
a
l
EC
G
a
n
a
l
y
s
i
s
v
i
a
g
r
a
mi
a
n
a
n
g
u
l
a
r
f
i
e
l
d
s
a
n
d
sp
l
i
t
a
t
t
e
n
t
i
o
n
,
”
i
n
N
e
u
ra
l
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
(
I
C
O
N
I
P
2
0
2
4
)
,
2
0
2
6
,
p
p
.
2
9
9
–
3
1
2
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
96
-
6
6
0
3
-
4
_
2
1
.
[
1
9
]
T.
S
u
b
b
a
a
n
d
T
.
C
h
i
n
g
t
h
a
m,
“
C
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
s
i
s
o
f
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
w
i
t
h
a
d
v
a
n
c
e
d
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
f
o
r
E
C
G
si
g
n
a
l
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
5
7
7
2
7
–
5
7
7
4
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
3
8
7
0
4
1
.
[
2
0
]
S
.
S
.
P
a
n
d
i
,
A
.
S
e
n
t
h
i
l
s
e
l
v
i
,
T.
K
u
marag
u
r
u
b
a
r
a
n
,
a
n
d
S
.
D
h
a
n
a
s
e
k
a
r
a
n
,
“
S
e
l
f
-
a
t
t
e
n
t
i
o
n
-
b
a
s
e
d
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
o
p
t
i
m
i
z
e
d
w
i
t
h
c
o
l
o
r
h
a
r
m
o
n
y
a
l
g
o
r
i
t
h
m
f
o
r
b
r
a
i
n
t
u
mo
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
E
l
e
c
t
r
o
m
a
g
n
e
t
i
c
B
i
o
l
o
g
y
a
n
d
M
e
d
i
c
i
n
e
,
v
o
l
.
4
3
,
n
o
.
1
–
2
,
p
p
.
3
1
–
4
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
8
0
/
1
5
3
6
8
3
7
8
.
2
0
2
4
.
2
3
1
2
3
6
3
.
[
2
1
]
I
.
G
o
p
a
l
a
k
r
i
s
h
n
a
n
,
H
.
M
a
h
a
r
a
j
a
n
,
V
.
S
a
n
k
a
r
,
V
.
B
h
a
r
a
n
i
k
u
mar,
a
n
d
S
.
O
l
i
v
e
r
A
,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
s
t
r
a
t
e
g
y
f
o
r
d
e
t
e
c
t
i
n
g
a
n
d
a
n
a
l
y
z
i
n
g
f
a
k
e
o
n
l
i
n
e
c
o
n
t
e
n
t
,
”
i
n
2
0
2
3
G
l
o
b
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
s
(
G
C
I
T
C
)
,
2
0
2
3
,
p
p
.
1
–
7
,
d
o
i
:
1
0
.
1
1
0
9
/
G
C
I
TC
6
0
4
0
6
.
2
0
2
3
.
1
0
4
2
6
4
3
3
.
[
2
2
]
S
.
Y
o
u
,
B
.
G
u
,
Y
.
Y
a
n
g
,
S
.
Y
u
,
a
n
d
S
.
G
u
o
,
“
D
T
4
E
C
G
:
a
d
u
a
l
-
t
a
s
k
l
e
a
r
n
i
n
g
f
r
a
m
e
w
o
r
k
f
o
r
EC
G
-
b
a
se
d
h
u
m
a
n
i
d
e
n
t
i
t
y
r
e
c
o
g
n
i
t
i
o
n
a
n
d
h
u
m
a
n
a
c
t
i
v
i
t
y
d
e
t
e
c
t
i
o
n
,
”
2
0
2
5
,
a
r
X
i
v
:
2
5
0
2
.
1
1
0
2
3
.
[
2
3
]
R
.
Z
h
u
,
G
.
-
Y
.
Z
h
o
n
g
,
a
n
d
J.
-
C
.
Li
,
“
F
o
r
e
c
a
s
t
i
n
g
p
r
i
c
e
i
n
a
n
e
w
h
y
b
r
i
d
n
e
u
r
a
l
n
e
t
w
o
r
k
mo
d
e
l
w
i
t
h
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
E
x
p
e
rt
S
y
s
t
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
4
9
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
4
.
1
2
3
6
9
7
.
[
2
4
]
C
.
Y
a
n
g
a
n
d
C
.
D
i
n
g
,
“
L
e
a
r
n
i
n
g
w
o
r
d
e
m
b
e
d
d
i
n
g
w
i
t
h
b
e
t
t
e
r
d
i
s
t
a
n
c
e
w
e
i
g
h
t
i
n
g
a
n
d
w
i
n
d
o
w
si
z
e
sc
h
e
d
u
l
i
n
g
,
”
2
0
2
4
,
a
r
X
i
v
:
2
4
0
4
.
1
4
6
3
1
.
[
2
5
]
K
.
C
h
h
a
p
a
r
i
y
a
,
K
.
M
.
B
u
d
d
h
i
r
a
j
u
,
a
n
d
A
.
K
u
mar,
“
A
d
e
e
p
s
p
e
c
t
r
a
l
–
s
p
a
t
i
a
l
r
e
si
d
u
a
l
a
t
t
e
n
t
i
o
n
n
e
t
w
o
r
k
f
o
r
h
y
p
e
r
sp
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
EEE
J
o
u
rn
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
Ap
p
l
i
e
d
E
a
rt
h
O
b
s
e
rv
a
t
i
o
n
s
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
1
7
,
p
p
.
1
5
3
9
3
–
1
5
4
0
6
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
2
4
.
3
3
5
5
0
7
1
.
[
2
6
]
M
.
W
u
,
Y
.
L
u
,
W
.
Y
a
n
g
,
a
n
d
S
.
Y
.
W
o
n
g
,
“
A
st
u
d
y
o
n
a
r
r
h
y
t
h
mi
a
v
i
a
E
C
G
si
g
n
a
l
c
l
a
ssi
f
i
c
a
t
i
o
n
u
s
i
n
g
t
h
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
Fr
o
n
t
i
e
rs
i
n
C
o
m
p
u
t
a
t
i
o
n
a
l
N
e
u
r
o
sc
i
e
n
c
e
,
v
o
l
.
1
4
,
2
0
2
1
,
d
o
i
:
1
0
.
3
3
8
9
/
f
n
c
o
m.
2
0
2
0
.
5
6
4
0
1
5
.
[
2
7
]
S
.
P
.
K
a
b
a
d
e
,
S
.
K
.
P
a
n
d
e
y
,
a
n
d
M
.
K
.
O
j
h
a
,
“
C
l
a
ssi
f
i
c
a
t
i
o
n
o
f
a
r
r
h
y
t
h
m
i
a
h
e
a
r
t
b
e
a
t
s
b
a
s
e
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
u
s
i
n
g
El
e
c
t
r
o
c
a
r
d
i
o
g
r
a
m
si
g
n
a
l
s
,
”
i
n
2
0
2
5
2
n
d
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
f
o
r
I
n
n
o
v
a
t
i
o
n
s
i
n
H
e
a
l
t
h
c
a
re
I
n
d
u
st
r
i
e
s
(
I
C
AI
I
H
I
)
,
2
0
2
5
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
I
I
H
I
6
7
1
2
4
.
2
0
2
5
.
1
1
4
0
3
7
9
8
.
[
2
8
]
S
.
K
.
P
a
n
d
e
y
e
t
a
l
.
,
“
D
e
t
e
c
t
i
o
n
o
f
a
r
r
h
y
t
h
m
i
a
h
e
a
r
t
b
e
a
t
s
f
r
o
m
EC
G
si
g
n
a
l
u
si
n
g
w
a
v
e
l
e
t
t
r
a
n
sf
o
r
m
-
b
a
se
d
C
N
N
m
o
d
e
l
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
S
y
s
t
e
m
s
,
v
o
l
.
1
6
,
n
o
.
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
4
1
9
6
-
0
2
3
-
0
0
2
5
6
-
z.
[
2
9
]
A
.
B
u
z
e
l
i
n
e
t
a
l
.
,
“
A
C
N
N
-
b
a
s
e
d
l
o
c
a
l
-
g
l
o
b
a
l
s
e
l
f
-
a
t
t
e
n
t
i
o
n
v
i
a
a
v
e
r
a
g
e
d
w
i
n
d
o
w
e
mb
e
d
d
i
n
g
s
f
o
r
h
i
e
r
a
r
c
h
i
c
a
l
EC
G
a
n
a
l
y
s
i
s,
”
i
n
Ma
c
h
i
n
e
L
e
a
r
n
i
n
g
a
n
d
K
n
o
w
l
e
d
g
e
D
i
sco
v
e
ry
i
n
D
a
t
a
b
a
s
e
s
,
2
0
2
6
,
p
p
.
2
8
3
–
2
9
9
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
2
-
0
6
0
6
6
-
2
_
1
7
.
[
3
0
]
K
.
Tz
e
v
e
l
e
k
a
k
i
s,
Z.
S
t
e
f
a
n
i
d
i
,
a
n
d
G
.
M
a
r
g
e
t
i
s,
“
R
e
a
l
-
t
i
me
st
r
e
ss
l
e
v
e
l
f
e
e
d
b
a
c
k
f
r
o
m
r
a
w
EC
G
si
g
n
a
l
s
f
o
r
p
e
r
s
o
n
a
l
i
se
d
,
c
o
n
t
e
x
t
-
a
w
a
r
e
a
p
p
l
i
c
a
t
i
o
n
s
u
s
i
n
g
l
i
g
h
t
w
e
i
g
h
t
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
a
r
c
h
i
t
e
c
t
u
r
e
s
,
”
S
e
n
so
rs
,
v
o
l
.
2
1
,
n
o
.
2
3
,
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
1
2
3
7
8
0
2
.
[
3
1
]
P
.
N
.
M
a
l
l
e
sw
a
r
i
,
V
.
k
r
i
s
h
n
a
O
d
u
g
u
,
T.
J.
V
.
S
.
R
a
o
,
a
n
d
T.
V
.
N
.
L.
A
s
w
i
n
i
,
“
D
e
e
p
l
e
a
r
n
i
n
g
-
a
ssi
s
t
e
d
a
r
r
h
y
t
h
mi
a
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
2
-
D
EC
G
sp
e
c
t
r
o
g
r
a
ms,
”
EU
RAS
I
P
J
o
u
rn
a
l
o
n
Ad
v
a
n
c
e
s
i
n
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
,
v
o
l
.
2
0
2
4
,
n
o
.
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s
1
3
6
3
4
-
024
-
0
1
1
9
7
-
1.
[
3
2
]
K
.
Ze
i
n
a
l
i
p
o
u
r
a
n
d
M
.
G
o
r
i
,
“
G
r
a
p
h
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
t
o
p
o
l
o
g
i
c
a
l
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
i
n
EC
G
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
in
A
p
p
l
i
c
a
t
i
o
n
s
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
N
e
u
r
a
l
S
y
st
e
m
s t
o
D
a
t
a
S
c
i
e
n
c
e
,
v
o
l
.
360
,
S
p
r
i
n
g
e
r
,
S
i
n
g
a
p
o
r
e
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
99
-
3
5
9
2
-
5
_
2
.
[
3
3
]
M
.
D
e
g
i
r
me
n
c
i
,
M
.
A
.
O
z
d
e
mi
r
,
E.
I
z
c
i
,
a
n
d
A
.
A
k
a
n
,
“
A
r
r
h
y
t
h
m
i
c
h
e
a
r
t
b
e
a
t
c
l
a
s
si
f
i
c
a
t
i
o
n
u
s
i
n
g
2
D
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
I
n
n
o
v
a
t
i
o
n
a
n
d
R
e
s
e
a
r
c
h
i
n
Bi
o
M
e
d
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
4
3
,
n
o
.
5
,
p
p
.
4
2
2
–
4
3
3
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
r
b
m
.
2
0
2
1
.
0
4
.
0
0
2
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
P.
K
a
v
ith
a
h
o
l
d
s
a
M
.
E.
d
e
g
r
e
e
in
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
fro
m
th
e
Un
iv
e
rsity
o
f
A
n
n
a
Un
iv
e
rsity
,
Ch
e
n
n
a
i
,
in
2
0
1
4
.
C
u
rre
n
tl
y
d
o
i
n
g
P
h
.
D.
(
p
a
rt
ti
m
e
)
in
P
re
sid
e
n
c
y
Un
i
v
e
rsity
,
Be
n
g
a
lu
r
u
In
d
ia.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
ima
g
e
p
ro
c
e
ss
in
g
,
m
a
c
h
in
e
lea
rn
in
g
,
a
n
d
d
e
e
p
lea
rn
in
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
k
a
v
it
h
a
.
2
0
2
3
3
CS
E
0
0
2
4
@
p
re
sid
e
n
c
y
u
n
iv
e
rsit
y
.
i
n
.
Dr
.
L.
S
h
a
k
k
e
e
r
a
re
c
e
iv
e
d
h
e
r
P
h
.
D.
d
e
g
re
e
fro
m
B.
S
.
A
b
d
u
r
Ra
h
m
a
n
Cre
sc
e
n
t
In
stit
u
te
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
An
n
a
Un
i
v
e
rsity
,
C
h
e
n
n
a
i,
In
d
ia
in
2
0
1
8
.
S
h
e
is
c
u
rre
n
tl
y
a
p
ro
fe
ss
o
r
a
n
d
a
ss
o
c
ia
te
d
e
a
n
-
a
c
a
d
e
m
ics
in
th
e
S
c
h
o
o
l
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
a
n
d
In
f
o
rm
a
ti
o
n
S
c
ie
n
c
e
fro
m
P
re
sid
e
n
c
y
Un
iv
e
rsit
y
,
Be
n
g
a
lu
r
u
,
In
d
ia.
He
r
a
re
a
s
o
f
re
se
a
rc
h
in
tere
st
in
c
lu
d
e
m
o
b
il
e
c
lo
u
d
c
o
m
p
u
t
in
g
,
m
a
c
h
in
e
lea
rn
in
g
,
I
o
T,
M
AN
E
T,
in
fo
rm
a
ti
o
n
se
c
u
ri
ty
,
a
n
d
d
a
ta
a
n
a
ly
t
ics
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sh
a
k
k
e
e
ra
.
l@p
re
sid
e
n
c
y
u
n
iv
e
rsit
y
.
in
.
Evaluation Warning : The document was created with Spire.PDF for Python.