Monitoring of Non-conventional abdominal myoelectrical ...

Non-conventional abdominal signals: EEnG & EHG

PhD. Javier García Casado

[email protected]

Monitoring of Non-conventional abdominal myoelectrical recordings: intestinal (EEnG) and uterine activity (EHG)

Non-conventional abdominal signals: EEnG & EHGNon-conventional abdominal signals: EEnG & EHG

Index

o Presentation

o Intestinal activity studies

Internal EEnG. Caracterization

Surface EEnG. • Interference cancellation: Software

• Interference cancellation: Hardware

o Uterine activity studies

Preterm birth

Uterine dynamics monitoring

Enhancing and characterization of EHG in pregancy and labor


Presentation

o Gbio-e tiene has more than 20 years of experience on Biomedical Engineering. Now is part of I3BH-UPV, Valencia

o Formed by 24 researchers (15 PhD)

o Principal Areas of R+D+I

Biomedical Systems and Devices for Diagnosis and Therapy

• Development of sensors and biosensors

• Development of Ad hoc monitoring systems

• Advanced signal processing and decisson support systems

Modelling and Simulation of Cells and Tissues

• Cardiac electrical activity models

• Signal simulation for diagnosis

o Collaborates with private companies, hospitals

and international research groups.



o Biolectrical signals on body surface

Electroencephalogram (EEG)

Electrooculogram (EOG)

Electrocardiogram (ECG)

Electrogastrogram (EGG)

Electrohysterogram (EHG)

Electromyogram (EMG)

Electroenterogram (EEnG)



Intestinal Activity Studies

o Intestinal Motility Segmentation and propelling the chyme

Mix food and enzymes

Gastrointestinal disorders are associated to intestinal motility dysfunctions

• Intestinal ischemia

• Intestinal obstruction

• Duodenal distention

• Myopathic disorders

• Paralytic ileum

• Thyroid abnormalities

• Etc.



o Pressure Recordings

o Myoelectrical Recordings

0 5 10 15 20 25 30time (s)

Slow Wave

IntestinalContractions

Spike Bursts Electroenterogram (EEnG)




0 5 10 15 20 25 30tiempo (s)


o Objetives Characterize internal EEnG

Relate it with Gold Standard (Pressure)



EE

nG (

mV

)

time(s)0 5 10 15 20 25 30

1.5

-1.5

Indi

ce M

otil

idad

(u.

a.)

200 6040 10080 140120 160 200180 220time (min)

240

Phase I: Inactivity 50min

Phase II: Irregular Activity 30min

Phase III: Maximum Activity 6min

I IIIII

Interdigestive Migratory Motor Complex (IMMC)

IMMC


o Activity Pattern in Fast State



EE

nG (

mV

)

time(s)0 5 10 15 20 25 30

1.5

-1.5

Indi

ce M

otil

idad

(u.

a.)

200 6040 10080 140120 160 200180 220time (min)

240




I IIIII


IMMC





EE

nG (

mV

)

time(s)0 5 10 15 20 25 30

1.5

-1.5

Indi

ce M

otil

idad

(u.

a.)

200 6040 10080 140120 160 200180 220time (min)

240




I IIIII


IMMC





Indi

ce M

otil

idad

(u.

a.)

200 6040 10080 140120 160 200180 220tiempo (min)

240

After breath intake till the return of Phase III

Irregular contractile activity

Duration and intensity depend on the quantity and composition of ingested food

PostingestionPatternIngest

o Activity Pattern in Postingestion




EE

nG (

mV

)

10 2 3 4-1.5

1.5

time (s)

Amplitude

● Amplitude SB

● Nº of SB/min● Duration SB

Duration

● Nº of peaks/min

Nº peaks


o Intestinal Motility Parameters



● Amplitude SB● Duration SB

time (s)0 5 10 15 20 25 30

1.5

-1.5

EE

nG (

mV

)

time (s)0 5 10 15 20 25 30

1.5

-1.5

EE

nG (

mV

)

Threshold levels: subjective criteria



0 20 40 60 80 100 120 140 160 180 200 220 240

Time (min)

NºS

B/m

in

● Does not include information about the INTENSITY of the contractions


?

● Nº of SB/min


0 20 40 60 80 100 120 140 160 180 200 220 240

Tiempo (min)

RM

S

● Root Mean Square (RMS)

Energy SW





EE

nG (

mV

)

10 2 3 4-1.5

1.5

time (s)● Slow Wave (SW)

■ Slow oscillation of potential

● Spike Bursts (SB)■ Rapid changes of potential

Low Frequency

High Frequeny

Multicomponent Signal




frecuencia (Hz)5.000.15 1.00 2.00 3.00 4.00

PSW

1.0

-1.0

tiempo (s)1200 10 20 30 40 50 60 70 80 90 100 110

(mV)InternalEEnG

● Decreasing amplitude peaks

● At multiples of the SW repetition frequency (0.3Hz,18 cpm dog)

o Spectral Analysis of the SW




frecuencia (Hz)5.000.15 1.00 2.00 3.00 4.00

POL


■ The application of this technique to monitor EEnG dominant frequency (freq. SW) can detect pathologies such as intestinal ischemia

Intestinal

ischemia

DF


● At multiples of the SW repetition frequency (0.3Hz,18 cpm dog)



frecuencia (Hz)5.000.15 1.00 2.00 3.00 4.00

PSW

● Up to what frequency does it extend?

100100

1

0

0

10

010

f

f

w

f

f

w

dffPT

dffPT

f,fE

f,fEf,f,fE%

Residual energy tail =2.5%

● Where is concentrated the energy of the SW? %ESW(f)

● %Energy accumulated in frequency of the SW:

fhcSW● High limit in freq. of SW:

100

0

20

40

60

80

frecuencia (Hz)5.000.15 1.00 2.00 3.00 4.00

%ESW

fhcSW

% 97.5%SW OL SW

hc hcf E f

1.950.60 Hz






frecuencia (Hz)35.000.15 1.00 10.00

PN

SB, P

SB

No SB - SB

contractile inactivity (No SB)1.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

maximum contractile activity (SB)1.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)



+Spike Bursts

SW ● Power spectral density in range SW >> amplitude than higher frequencies



No SB - SB

● Power spectral density above 2 Hz increases significantly during periods of contractile activity


-1.0

tiempo (s)300 5 10 15 20 25

(mV)


-1.0

tiempo (s)300 5 10 15 20 25

(mV)



+Spike Bursts

frecuencia (Hz)35.002.00 10.00

PN

SB, P

SB


frecuencia (Hz)35.002.00 10.00

PN

SB, P

SB

o Spectral Analysis of the SW+Spike Bursts

● Energy Limits of SB?


-1.0

tiempo (s)300 5 10 15 20 25

(mV)


-1.0

tiempo (s)300 5 10 15 20 25

(mV)


o Increment in signal energy: contractile activity-inactivity

o Higher frequencies than SW range: f>2 Hz100

0

20

40

60

80

frecuencia (Hz)352 10

%E

SB

-NS

B

fhcSB

flcSB

27,251.30 Hz2,250.07 Hz



RM

S>

2Hz

RMS Orig.RMS FPA2Hz

0 20 40 60 80 100 120 140 160 180 200 220 240

Tiempo (min)

P2H

z^0.

5

1

RMS Orig.P2Hz0.5




removedSW energy

removedSW energy

High Pass Filter

f

G

2Hz

X(t) Y(t)

SpectralEstimator

X(t) P(f)




IM Pressure

IM EEnG



9.000.15 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00

PS

D (

mV

2 /H

z)

frecquency (Hz) 30

o Spectral Analysis Requires signals to be stationary

Lost of temporal information

Solution: Time-Frequency Analysis





o Time-Frequency Analysis: Permits independent resolution for T and F

The kernel function determines the properties of the distribution

Time-Frequency Analysis


T-F generalized (Cohen)


Intestinalparalisis


o Time-Frequency Analysis: Similarly to quantifying parameters from PSD,

marginals in T and F can be defined

Marginals

Detection of Gastrointestinal Pathologies



o Internal Myoelectrical RecordingsRequire surgical intervention

No clinical application

Sensor

o Surface Myoelectrical RecordingsNon Invasive

Low Cost


o Objetives Characterize Surface EEnG

Relate it with Internal EEnG



o Internal Myoelectrical RecordingsRequire surgical intervention

No clinical application

o Surface Myoelectrical RecordingsNon Invasive

Low Cost


o Objetives Characterize Surface EEnG

Relate it with Internal EEnG



o Surface EEnG Very attenuated signal, specially in ↑freq. (SB)contractile inactivity (No SB)

0.1

-0.11.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

(mV)

SurfaceEEnG

InternalEEnG


-0.11.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

(mV)

SurfaceEEnG

InternalEEnG




o Surface EEnG Very attenuated signal, specially in ↑freq. (SB)


SurfaceRecording

AbdominalLayers

InternalRecording

1.0

-1.0

(mV)

0.1

-0.11 0

(mV)

● Estimation of the transfer function of abdominal layers for each freq.■ Energy ratio surface / internal in bands of 1 Hz

Gabd(f)su

int

( 1, )

( 1, )

SBrf

SB

E f f

E f f

for f=1,2,…35 Hz

Gabd(f)Eint(f) Esurf(f)



o Surface EEnG Very attenuated signal, specially in ↑freq. (SB)


-15

-35

-30

-25

-20

frecuencia (Hz)35.00.1 1.0 10.0

Gab

d(d

B)

● 1st segment: f → higher attenuation

Low pass filter of EEnG energy

● 2nd segment: f → lower attenuation

aditional energy in surface EEnG?

Most studies focus on the SW (low frequency)

in high frequencyINTERFERENCE


SurfaceRecording

AbdominalLayers

InternalRecording

Gabd(f)Eint(f) Esurf(f)

+Einterf



0.1

-0.11.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

(mV)

SurfaceEEnG

InternalEEnG


-0.11.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

(mV)

SurfaceEEnG

InternalEEnG

Presents interferences:○ Contact potential

○ EGG

○ Respiration

Cancellation?


○ ECG

○ Movement artifacts

Low Freq. High Freq.‘Software’ ‘Hardware’



Refer.Interf.

S+IECG

ÎECG

S


o Interference Cancellation: ‘Software’ Adaptive Filtering ECG

Cancellation?

‘Software’



Refer.Interf.

S+IECG

ÎECG

S



Tp Ta

0.08

-0.08

Time (s)100 1 2 3 4 5 6 7 8 9

0.70

-0.70100 1 2 3 4 5 6 7 8 9

0.10

-0.10100 1 2 3 4 5 6 7 8 9

EC

G (

mV

)E

EnG

(m

V)

Inte

rf. (

mV

)

n=1

Tp Ta0

n=13

Tp Ta0





Original

Adapt. Filt.

Original

Adapt. Filt.

SW

SW+SB





Surface

Internal

Effect on intestinal motility parameters


removedECG energy


● It’s a subgroup (the +powerful and +employed) of the Blind Source Separation (BSS) techniques

o Interference Cancellation: ‘Software’ Independent Component Analysis (ICA)




x t A·s t n t Observed Signals

Unknwown Mixing

Process (A)

s1

sn-1

sn

x1

xm

Source Signals


● Consists on extracting a set of statiscally independent components from a set of observed signals without prior knowlegde of the signal sources and the mixing matrix.





x t A·s t n t

ObservedEEnGUnknwown

Mixing Process

(A)

EEnG

Respiration

ECGMovem. Artif.







Unknwown Mixing

Process (A)

ICA

(A-1)

x t A·s t n t

ObservedEEnG

EEnG

Respiration

ECGMovem. Artif.

x1

x2







0 0.2 0.4 0.6 0.8 1 0 20 40 60Tiempo (s) Frecuencia (Hz)

0.05

5

-0.05 0.015

-0.015

a)

b)

g)

h)

x 1[k

]

(mV

)

PS

D

(a.u

.)

PS

D

(a.u

.)

x 2[k

]

(mV

)

0.055

-0.05 0.015

-0.015

a)

b)

c)

d)

g)

h)

i) 5

-5 5

-5

x 1[k

]

(mV

)

j)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

x 2[k

]

(mV

) s 1

[k]

(a

.u.)

s 2

[k]

(a

.u.)


■ ICA concentrates the components from the signal of interest and from interferences in different outputs (sources)

Remaining Interf.

Remaining signal

Respiration

Resp.

SW

slow wave


x1

x2



SW+interf. ↓f

slow waveICA out



0.1

-0.1 0.02

-0.02

a)

b)

g)

h)

x 1[k

] (m

V)

PSD

(a

.u.)

PSD

(a

.u.)

x 2[k

] (m

V)

0.1

-0.1 0.02

-0.02

a)

b)

c)

d)

g)

h)

i) 0.05

-0.05 0.05

-0.05

x 1[k

] (m

V)

)

j)

PSD

(a

.u.)

PSD

(a

.u.)

PSD

(a

.u.)

PS

D

(a.u

.)

x 2[k

] (m

V)

0 01

s 1[k

] (a

.u.)

s 2[k

] (a

.u.)


Remaining signal

Interf.↓↓f (gastric)

Interf.

SW


■ ICA concentrates the components from the signal of interest and from interferences in different outputs (sources)

x1

x2



slow wave

slow wave

ICA out

Remaining Interf.

SW+interf. ↓f

Non-conventional abdominal signals: EEnG & EHGNon-conventional abdominal signals: EEnG & EHGNon-conventional abdominal signals: EEnG & EHG


20 40 60 80 1000

0.1

0.2

0.3

0.4

0.5

0

0.1

0.2

0.3

0.4

0.5 D

F (

Hz)

D

F (

Hz)

b)

Tiempo (min)

a)

120 140 0

Channel 1

Channel 2

x Original○ Processed

?


0.05

-0.05 0.01

-0.01

a)

b)

g)

h)

5

x 1[k

]

(mV

)

PS

D

(a.u

.)

PS

D

(a.u

.)

x 2[k

]

(mV

)


0.05

-0.05 0.01

-0.01

a)

b)

c)

d)

g)

h)

i) 5

-5 5

-5

x 1[k

]

(mV

)

j)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

x 2[k

]

(mV

) s 1

[k]

(a

.u.)

s 2

[k]

(a

.u.)


Interf.↓↓f

Respiration

■ If there is more than 1 source of interference, ICA cannot separate the sources correctly.

Nº Chanels must be greater than Nº sources


x1

x2


slow wave

Remain.Interf.

SW+interf.

ICA out

SW+interf. ↓f


1

n

j nj

x t IMF t r t

o Interference Cancellation: ‘Software’ Empirical Mode Decomposition (EMD)

● The Empirical Mode Decomposition (EMD) consists on decomposing a signal or temporal series in a finite number of Intrinsic Mode Functions (IMF)

● The IMF are non linear oscillatory functions that are directly extracted from the data. They must satisfy 2 conditions:■ the nº of extrema and the nº of zero crossing differ by one at most■ the average of the upper and lower envelope must be sufficiently close to

zero according to some criterion

● The decomposition process is adaptive and data driven. Empiric method, NO explicit equations.

● The technique is applicable to non-stationary and non-linear signals






-0.015

0.015

x[k]

(m

V)

g) a)

PS

D

(a.u

.)

SIR=0.3 dB

0 0.2 0.4 0.6 0.8 1 0 20 40 60 Tiempo (s) Frecuencia (Hz)


-0.015

0.015 -0.015

0.015

-0.015

0.015 -0.015

0.015

x[k]

(m

V)

IMF

1[k]

(m

V)

r 3 [

k]

(mV

) y[

k] (

mV

)

g)

h)

i)

j)

k)

a)

b)

c)

d)

e)

f)

-0.015

0.015

-0.015

0.015

IMF

2[k]

(m

V)

IMF

3[k]

(m

V)

l)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

SIR=18.6 dB

SIR=0.3 dB

Resp.

Resp.



EMD out

Original

Processed

Resp.

SW+interf. ↓f


-0.1

0.1 g) a)

x[k]

(m

V)

PS

D

(a.u

.)

SIR=-4 dB



-0.1

0.1

-0.1

0.1

-0.1

0.1

-0.1

0.1 -0.1

0.1 g)

h)

i)

j)

k)

l)

a)

b)

c)

d)

e)

f)

-0.1

0.1

x[k]

(m

V)

IMF

1[k]

(m

V)

r 3 [

k]

(mV

) y[

k] (

mV

) IM

F2[

k]

(mV

) IM

F3[

k]

(mV

)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

SIR=17.9 dB

SIR=-4 dB


Interf. ↓↓f




EMD out

Processed

SW+interf. ↓f




-0.05

0.05

x[k]

(m

V)

g) a)

PS

D

(a.u

.)

SIR=2.8 dB

0 0.2 0.4 0.6 0.8 1 0 20 40 60 Tiempo (s) Frecuencia (Hz)-0.05

0.05 -0.05

0.05

-0.05

0.05 -0.05

0.05

x[k]

(m

V)

IMF

1[k]

(m

V)

r 3 [

k]

(mV

) y[

k] (

mV

)

g)

h)

i)

j)

k)

a)

b)

c)

d)

e)

f)

-0.05

0.05

-0.05

0.05

IMF

2[k]

(m

V)

IMF

3[k]

(m

V)

l)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

PS

D

(a.u

.)

SIR=2.8 dB

SIR=13.8 dB

Interf. ↓↓f

Interf.↓↓f

Resp.

Resp.


EMD out

Processed

SW+interf. ↓f

DF



0

0.5

0.4

0.3

0.2

0.1

0

DF

(H

z)

Tiempo (min)20 80 140 100 40 60 120

x Original○ Processed


SW+interf. ↓f

DF


0

0.5

0

0.50

0.50

0.50

0.50

0.5

DF2

(Hz)

DF2

(Hz)

DF2

(Hz)

DF2

(Hz)

DF2

(Hz)

DF2

(Hz)

Time (min)0 50 100 150 200 250

0.25±0.03 Hz

0.26±0.03 Hz

0.25±0.03 Hz

0.25±0.03 Hz

0.23±0.02 Hz

0.21±0.02 Hz

a)

b)

c)

d)

e)

f)



DF Surface

DF Internal 1

DF Internal 2

DF Internal 3

DF Internal 4

DF Internal 5

SW+interf. ↓f



0.1

-0.11.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

(mV)

SurfaceEEnG

InternalEEnG


-0.11.0

-1.0

tiempo (s)300 5 10 15 20 25

(mV)

(mV)

SurfaceEEnG

InternalEEnG

Presents interferences:○ Contact potential

○ EGG

○ Respiration

Cancellation?


○ ECG

○ Movement artifacts

Low Freq. High Freq.‘Software’ ‘Hardware’



o Interference Cancellation: ‘Hardware’ Electrodes in Laplacian configuration

Cancellation

‘Hardware’



1

2

3

4 5

Rigid Electrodes


5 1 2 3 42

4 1

4VL V V V V Vb


Flexible Electrodes

Discrete approximation Continuous approximation




Rigid Electrodes


0

3

PS

D (

a.u) Resp

0

4e-3

PS

D (

a.u) Ch1

0 10 20 300

5e-4

PS

D (

a.u) TCB

Frequency (cpm)

-2

2

(V)

Resp

-0.1

0.1

(mV

)

Ch1

0 60-1

1

(mV

) ECG

Time (seconds)

(mV

)

-0.1

0.1

(mV

) TCB

(mV

)

Respiration

EEnG

Bipolar

EEnG

Laplacian

ECG

-2

2

(V)

Resp

-0.1

0.1Ch1

0 10-1

1

(mV

)

ECG

Time (seconds)

-0.1

0.1TCB

SW


Interf. ECG ↓

Int. LF↓ Int. Resp.↓




Int. Resp.↓



Uterine Activity Studies

o Preterm birth (<37 wog) 5-10% of total births

Main cause of perinatal deaths (85%); 1/5 mental retardation; 1/3 visual impairment; 1/2 cerebral palsy

73

54

41

32

9

63

85

82

85

23

24

25

26

27Ges

tati

on

alA

ge(

wee

ks)

0 20 40 60 80 100(%)Mortality %Survival without moderate or severe sequelae





It involves prolonged hospital stays and a high economic cost

Semanas

36343230282624

Cos

te d

e cu

idad

os (

x 10

00 S

)

120

100

80

60

40

20

0

Car

eC

ost

(x10

00 $

)Weeks of gestation

Cost of surviving newborn

96

90

54

43

31

66

38

10

7

4

500 - 749

750 - 999

1000-1249

1250-1499

1500-1999

Newborn weight (gr)

0 20 40 60 80 100 120Average stay (days)

TotalIntens. Care

Excluding deaths and readmissions





It involves prolonged hospital stays and a high economic cost

The effectiveness of tocolytic and lung madurationagents is subject to the early initiation of therapy

Reliable prediction of true labor

understanding the mechanisms that initiate labor

study of uterine contractions.


o Monitoring of Uterine Contractions Intrauterine Pressure (IUP)

• Direct and accurate measurement

• ‘Free’ of artifacts and interferences

• ‘Only’ mechanical information

• Requires membrane rupture

Gold Standard



o Monitoring of Uterine Contractions External Tocography (TOCO)

• Non invasive measurement

• Provides approximate duration and frequency of contractions

• Inaccurate measurement– Influence of probe location and pressure

– Subjectivity of the examiner

• Use of tight straps

• No information about efficiency of the contractions



o Monitoring of Uterine Contractions

• Non invasive measurement

• Provides approximate duration and frequency of contractions

• Inaccurate measurement

• Use of tight straps

• No information about efficiency of the contractions

?


Electrohysterography (EHG)

•is NOT needed

– Amplitude depends on recording conditions

– Influence of body fat (less than TOCO)


-0.5

0

0.5

Can

al 3

(mV

)

-0.5

0

0.5

Can

al 4

(mV

)

-0.5

0

0.5

Can

al 3

(mV

)

-0.5

0

0.5

Can

al 4

(mV

)

-0.4

0

0.4

Bip

olar

(mV

)

0 20 40 60 80 100-0.4

0

0.4

Time (s)

Bip

olar

filt

rado

(mV

)

• Monopolar• Monopolar vs Bipolar


o EHG Recordings

Presents interferences from:○ Contact potential

○ Respiration

Cancellation?

○ ECG

○ Movt. Artifacts

‘Software’

‘Hardware’


A

f (Hz)351

Preprocessing

o Interference Cancellation: ‘Software’ ECG cancellation by ICA


1

16

x1[k]

x16[k]

.

.

.

ICA

G(s)=log(cosh(s))

IC1[k]

IC16[k]

.

.

.

(z)MxK

Identification & cancelling cardiac

ICs

Reconstruction Observed Signals


o Interference Cancellation: ‘Software’ ECG cancellation by ICA


0 50 100 150 200-0.2

0

0.2

Tiempo (s)

EC

G

(mV

)

-0.2

0

0.2

ICA

(m

V)

-0.2

0

0.2

Filt

rado

(mV

)

-0.2

0

0.2

Ori

gina

l(m

V)

(a)

(b)

(c)

(d)

30 40 50 60 70 80 90 100 110 120-0.2

0

0.2

Tiempo (s)

-0.2

0

0.2-0.2

0

0.2-0.2

0

0.2(a)

(b)

(d)

(c)

0 5 10 15 20-0.2

0

0.2

Tiempo (s)

-0.2

0

0.2-0.2

0

0.2-0.2

0

0.2(a)

(b)

(d)

(c)


o Interference Cancellation: ‘Software’ ECG cancellation by Adaptive Filtering


-0.05

0

0.05

Ch1

(m

V)

-0.05

0

0.05C

h2

(mV

)

-0.05

0

0.05

Ch3

(m

V)

-0.05

0

0.05

Ch4

(m

V)

0 1 2 3 4 5 6 7 8 9 10-0.2

0

0.2

EC

G m

ater

no(m

V)

Time (s)

Maternal ECG

Maternal+FetalECG


-0.05

0

0.05

Ch1

(m

V)

-0.05

0

0.05

Ch2

(m

V)

-0.05

0

0.05

Ch3

(m

V)

-0.05

0

0.05

Ch4

(m

V)

0 1 2 3 4 5 6 7 8 9 10-0.2

0

0.2

EC

G m

ater

no(m

V)

Time (s)

-0.05

0

0.05

Ch1

(m

V)

-0.05

0

0.05

Ch2

(m

V)

-0.05

0

0.05

Ch3

(m

V)

-0.05

0

0.05

Ch4

(m

V)

0 1 2 3 4 5 6 7 8 9 10-0.2

0

0.2

EC

G m

ater

no(m

V)

Time (s)

o Interference Cancellation: ‘Software’ ECG cancellation by Adaptive Filtering


Maternal ECG

FetalECG


0

50

100

IUP

(m

mH

g)

0 200 400 600 800 10000

50

100

Time (s)

TO

CO

(m

mH

g)

(d)

(c) -1

0

1

EH

G

(mV

)

(a)

o Interference Cancellation: ‘Software’ Artif. cancel. & contraction detection by expert system


N=645 Sensitivity Specificity PPV NPV

LDA 69.4% 97.0% 93.7% 83.1%

QDA 84.3% 97.0% 94.8% 90.6%

SVM 87.1% 97.3% 95.5% 92.1%

Toco-Like signal generationUterine ContractionsArtifactsIdentify & Classify



o SignalsEHG

• 5 Monopolar

• 4 Bipolar

• 1 Laplacian Discrete

• 2 Laplacian Continuous

M1

M2

M3

M4 M5

umbilicus

B1

B2

B4B4

L1 L2

LD

o Contraction detection: monitoring uterine dynamicscontraction detection

Cancellation?‘Software’

‘Hardware’



o SignalsEHG

• 5 Monopolar

• 4 Bipolar

• 1 Laplacian Discrete

• 2 Laplacian Continuous

TOCO

IUP

TOCO

IUP

o Contraction detection: monitoring uterine dynamics




1

2

C

T E

NCCI

N N

• NC nº consitent contract.• NT nº contract. IUP• NE nº contract. EHG (o toco)

S/IECG (dB) S/N (dB)

Cancellation?‘Hardware’




1

2

C

T E

NCCI

N N

• NC nº consitent contract.• NT nº contract. IUP• NE nº contract. EHG (o toco)

S/IECG (dB) S/N (dB)

Cancellation?‘Hardware’


• MPTLS=multiple preterm labor symptons • Ctx=Nºcontractions (TOCO)

• BS= Bishop Score

• Cx=cervical length (ultrasound)

•CFFN+= Cervical fetal fibronectin test

• LIF: Cervical light-induced fluorescence

• EMG= No-Invasive Uterine Electromiogram (EHG)

[Garfield & Maner, 2007]


o Preterm diagnosis


o Evolution throughout pregnancy

Physiological changes in EHG parameters:

• Amplitude

• Duration

• nº CT/h: 0.5‐6 CT/h pregancy vs. maximum 18 CT/h labor)

• Energy content shifts to higher frequencies

• relative energy of FWH



• Energy content shifts to higher frequencies

• relative energy of FWH

Mea

n pe

ak fr

eque

ncy


o Preterm diagnosis

[Terrien, 2009]

FWH swifts to ↑ frequencies


Preterm laborTerm labor


o Preterm diagnosis Dominant Frequency between [0.34,1] Hz [Manner, 2003]



o Preterm diagnosis Dominant Frequency between [0.34,1] Hz [Manner, 2003]



o Preterm diagnosis Dominant Frequency [Lukovnik, 2011]+ Propagation velocity



o Preterm diagnosis Dominant Frequency + Propagation velocity

Depends on electrode orientation and direction of propagated wavefrontNOT

V1V2



o Predicting Horizon and Mode of Delivery

PREGNANT:EHG Recording

LABOR

NON LABOR

Classifier: QDA, SVM(Accuracy 94%)



o Predicting Horizon and Mode of Delivery

PREGNANT:EHG Recording

LABOR

NON LABOR

SPONTANEOUS

INDUCTION /RIPENING

CESAREA

Mode of future delivery

TIME TO DELIVERY

(days)Classifier: QDA, SVM

(Accuracy 94%) Classifier(Accuracy 95%)

Estimator(error <1 day)

+ Obstetric Parameters



o Future work Enlarge databases

Study the effect of different cervical ripening drugs

Individualized follow up of EHG during pregnancy

Analyze propagation patterns and velocities

Develop portable EHG, fECG monitoring device

Aid to predict preterm labor


Dr. F. Javier García Casado

[email protected]

Thanks for your attention

Monitoring of Non-conventional abdominal myoelectrical ...

Documents

Transcript of Monitoring of Non-conventional abdominal myoelectrical ...