ORIGINAL RESEARCH
published: 16 May 2017
doi: 10.3389/ phys.2017.00305
F on ie s in Physiology | www. on ie sin.o g 1May 2017 | Volume 8 | A icle 305
Edi ed by:
Ahsan H. Khandoke ,
Khali a Uni e si y, UAE
Re iewed by:
Hai ham Alanga i,
Khali a Uni e si y, UAE
Faezeh Ma zban ad,
Monash Uni e si y, Aus alia
Yoshi aka Kimu a,
Tohoku Uni e si y, Japan
*Co espondence:
Janusz W obel
[email p o ec ed]
Special y sec ion:
This a icle was submi ed o
Compu a ional Physiology and
Medicine,
a sec ion o he jou nal
F on ie s in Physiology
Recei ed: 20 Ma ch 2017
Accep ed: 27 Ap il 2017
Published: 16 May 2017
Ci a ion:
Jezewski J, W obel J, Ma onia A,
Ho oba K, Ma inek R, Kupka T and
Jezewski M (2017) Is Abdominal Fe al
Elec oca diog aphy an Al e na i e o
Dopple Ul asound o FHR Va iabili y
E alua ion? F on . Physiol. 8:305.
doi: 10.3389/ phys.2017.00305
Is Abdominal Fe al
Elec oca diog aphy an Al e na i e o
Dopple Ul asound o FHR
Va iabili y E alua ion?
Janusz Jezewski1, Janusz W obel1*, Adam Ma onia1, K zysz o Ho oba1,
Radek Ma inek2, Tomasz Kupka1and Michal Jezewski3
1Ins i u e o Medical Technology and Equipmen ITAM, Zab ze, Poland, 2Depa men o Cybe ne ics and Biomedical
Enginee ing, VSB-Technical Uni e si y o Os a a, Os a a, Czechia, 3Ins i u e o Elec onics, Silesian Uni e si y o
Technology, Gliwice, Poland
G ea expec a ions a e connec ed wi h applica ion o indi ec e al elec oca diog aphy
(FECG), especially o home elemoni o ing o p egnancy. E alua ion o e al hea a e
(FHR) a iabili y, when de e mined om FECG, uses he same c i e ia as o FHR
signal acqui ed classically— h ough ul asound Dopple me hod (US). The e o e, he
equi alence o hose wo me hods has o be con i med, bo h in e ms o ecognizing
classical FHR pa e ns: baseline, accele a ions/decele a ions (A/D), long- e m a iabili y
(LTV), as well as e alua ing he FHR a iabili y wi h bea - o-bea accu acy—sho - e m
a iabili y (STV). The esea ch ma e ial consis ed o eco dings collec ed om 60 pa ien s
in physiological and complica ed p egnancy. The FHR signals o a leas 30 min du a ion
we e acqui ed dually, using wo sys ems o e al and ma e nal moni o ing, based on
US and FECG me hods. Reco dings we e e ospec i ely di ided in o no mal (41) and
abno mal (19) e al ou come. The complex p ocess o da a synch oniza ion and alida ion
was pe o med. Ob ained low le el o he signal loss (4.5% o US and 1.8% o FECG
me hod) enabled o pe o m bo h di ec compa ison o FHR signals, as well as indi ec
one—by using clinically ele an pa ame e s. Di ec compa ison showed ha he e is
no measu emen bias be ween he acquisi ion me hods, whe eas he mean absolu e
di e ence, impo an o bo h isual and compu e -aided signal analysis, was equal o
1.2 bpm. Such low di e ences do no a ec he isual assessmen o he FHR signal.
Howe e , in he indi ec compa ison he inconsis encies o se e al pe cen we e no ed.
This mainly a ec s he accele a ion (7.8%) and pa icula ly decele a ion (54%) pa e ns.
In he signals acqui ed using he elec oca diog aphy he ob ained STV and LTV indices
ha e shown signi ican o e es ima ion by 10 and 50% espec i ely. I also u ned ou ,
ha abili y o clinical pa ame e s o dis inguish be ween no mal and abno mal g oups do
no depend on he acquisi ion me hod. The ob ained esul s p o e ha he abdominal
FECG, conside ed as an al e na i e o he ul asound app oach, does no change he
in e p e a ion o he FHR signal, which was con i med du ing bo h isual assessmen
and au oma ed analysis.
Keywo ds: Dopple ul asound, e al elec oca diog am, e al hea a e analysis, e al s a e assessmen , e al
ou come
Jezewski e al. FHR om: aFECG s. US
INTRODUCTION
Fe al hea ac i i y is a p ima y sou ce o in o ma ion which
enables assessmen o he e al s a e du ing p egnancy and a
labo . This in o ma ion is ob ained mainly h ough analysis
o he e al hea a e (FHR) signal being o med om he
ins an aneous alues calcula ed acco ding o he o mula: FHR
[bpm] =60000/T [ms]. The FHR alues a e exp essed in
bea s pe minu e, and T is he ime in e al be ween wo
consecu i e e al hea bea s ha comp ises one comple e ca diac
cycle. Toge he wi h addi ional signals desc ibing he u e ine
con ac ile ac i i y and e al mo emen p o ile, he FHR signal
cons i u es he ca dio ocog aphic eco d. Acquisi ion o hese
addi ional signals is qui e simple, bu measu emen o he
e al hea a e has been always a challenge. Al eady in 1960s
he e al elec oca diog am was eco ded o he i s ime by
means o elec ode a ached o e al head. The quali y o such
eco ded di ec e al elec oca diog am (FECG) is usually e y
good, and hus i enables, using a qui e simple p ocessing
me hod, de e mina ion o he bea - o-bea in e als wi h e y
high accu acy. Howe e , he in asi e app oach and applica ion
limi ed o he labo only caused ha di ec me hod did no ound
wide applica ion in clinical p ac ice.
As a esul o u he esea ch and de elopmen he
nonin asi e ul asound (US) me hod has become a s anda d
app oach since ea ly 1970s, as i can be used bo h du ing
p egnancy and labo . A p esen , all bedside e al moni o s
in ended o use in clinical condi ions a e based on he pulsed
Dopple ul asound echnique, wi h measu emen ansduce
a ached o ma e nal abdomen. P inciple o ope a ion elies on
in e nal p ocessing o he en elope o he US beam e lec ed om
mo ing pa s o e al hea s— al es o walls, o ind he episodes
co esponding o consecu i e hea bea s. Howe e , a complex
s uc u e and a ying con en o he US signal, usually caused by
eloca ion o he e al hea in ela ion o a ansduce du ing
moni o ing session, make a de e mina ion o he bea - o-bea
in e al e y di icul (Khandoke e al., 2009; Ma zban ad e al.,
2014). The e o e, a co ela ion echniques, conside ing ull shape
o he analyzed signal, ha e been applied. The c oss-co ela ion
echnique wi h changeable empla e appea ed o be oo sensi i e
o US signal changes, which esul ed in conside able signals
loss. Thus, an au oco ela ion unc ion wi h adap i e window
selec ion has been applied in nex gene a ion e al moni o s.
Howe e , he au oco ela ion unc ion does no de ec he
consecu i e hea bea s bu only de e mines he ins an aneous
pe iodici y o he US signal en elope which co esponds o
ca diac cycle being measu ed. This leads o e ec o a e aging
o neighbo ing ca diac cycles and hus dec easing o FHR
de e mina ion accu acy in ela ion o e al elec oca diog aphy
(Lee e al., 2009; Voicu e al., 2010). The ob ained FHR signal
is p o ided by he bedside moni o as he ace in a p in ou
wi h es ablished ime scale o 1, 2, o 3 cm/min. As long as he
FHR ace has been analyzed isually, he lowe accu acy did
no a ec signi ican ly he e al s a e assessmen . Mo e impo an
was o ensu e he ace con inui y which allowed clinicians o
obse e a gene al endency o he e al hea a e changes, and
o ecognize he ea u es ep esen ing longi udinal FHR pa e ns
ela ing o he e al s a e, like accele a ion o decele a ion. I was
ound ha he e alua ion o e al s a e, when based on isual
in e p e a ion, has been mainly a ec ed by low in e - and in a-
obse e ag eemen (Jezewski e al., 2002; Romano e al., 2016a).
Tha was a esul o bo h complexi y o he FHR signal and he
ac ha impo an pa o in o ma ion ela ing o ins an aneous
changes o FHR alues has been hidden om a naked eye. These
changes a e conside ed o be e y impo an FHR cha ac e is ics,
e lec ing app op ia e neu ological modula ion o he FHR.
Thus, u he de elopmen s age o e al moni o ing was
aimed a au oma ed analysis o he FHR signal and i s
implemen a ion as buil -in p ocedu e o bedside moni o s as
well as in compu e -aided e al moni o ing sys em. Some o he
equi emen s impo an o moni o ing he p egnan women in
hospi al, like su eillance o many pa ien s, de ec ing and ale ing
o symp oms o e al dis ess, o elec onic a chi e wi h he
signals and pe ina al da a, ha e made he compu e -aided sys em
wi h online au oma ed analysis he s anda d in mode n obs e ics
(W obel e al., 2013, 2015a). Au oma ed analysis comp ises
de ec ion and desc ip ion o he abo e men ioned FHR ea u es,
like accele a ion and o he s, as well as de e mina ion o he
ins an aneous FHR changes by p o iding a se o indices
o e alua e he long- e m and sho - e m (bea - o-bea le el)
a iabili y o he e al hea a e.
Au oma ed online analysis p o ides a quan i a i e desc ip ion
o he FHR, bu he inal in e p e a ion o he eco d is s ill done
by a clinician. The e a e a numbe o pape s ela ing o au oma ed
classi ica ion o he FHR eco dings by using di e en me hods
o compu a ional in elligence like neu al ne wo k, suppo ec o
machines o epsilon-insensi i e lea ning (Czabanski e al., 2008,
2013). Howe e , aking in o accoun ha he inpu da a se
comp ised he au oma ically de e mined ea u es o he FHR
signals, collec ed om he clinical da abases (Chudacek e al.,
2014), he ob ained classi ica ion esul s should be aced o he
limi a ions o he ul asound app oach as i is discussed below
(Voicu e al., 2014; W obel e al., 2015c).
The a iabili y indices we e o iginally de ined using he bea -
o-bea in e als de e mined om he di ec elec oca diog am.
Thei s aigh applica ion o he FHR signals being p o ided
by e al moni o s aised a ques ion how a limi ed accu acy o
he ul asound app oach a ec s de e mina ion o he ca diac
in e als, and hus he a iabili y indices alues. Se e al esea ch
s udies we e aimed a e alua ion o he eliabili y o he
ul asound me hod in e e ence o he di ec elec oca diog aphy
(Ib ahimy e al., 2003; Reinha d e al., 2010, 2012; Cohen e al.,
2012; Kimu a e al., 2012; Desai e al., 2013; Ko ding e al.,
2015). Howe e , hey we e aimed a compa ing he signal loss
episodes o di ec ly he FHR alues. In ou s udy we showed
ha he e o o ca diac cycle de e mina ion (ins an aneous FHR
alue) has no been co ela ed wi h he FHR a iabili y indices
e o . I means ha he measu emen accu acy esul ing om
he e al moni o speci ica ion canno be di ec ly ela ed o he
esul s o he compu e -aided analysis o he FHR a iabili y.
In gene al, we concluded ha mode n e al moni o s using
he Dopple US echnique a e no able o p o ide he signal
wi h he accu acy equi ed o eliable quan i a i e e alua ion
o ins an aneous FHR a iabili y, pa icula ly he sho - e m
F on ie s in Physiology | www. on ie sin.o g 2May 2017 | Volume 8 | A icle 305
Jezewski e al. FHR om: aFECG s. US
a iabili y, based on he indices calcula ed au oma ically (Voicu
e al., 2014). Fo una ely, he alues o indices de e mined in ha
way a e unde es ima ed, which p e en s he e al dis ess signs
om being unde ec ed.
Se e al a emp s we e ca ied ou o imp o e he eliabili y o
he ul asound me hod by using ad anced signal p ocessing o
he Dopple en elope (Jezewski e al., 2011), bu none o hem
ha e been applied in he bedside moni o s ye . In W obel e al.
(2008) he me hod has been p oposed o imp o e he eliabili y o
he FHR a iabili y indices, which elies on he e o s ecognized
in he ul asound measu emen channel.
The ac , ha he FHR signal ob ained om he ul asound
app oach has been ecognized as no good enough o ully
exploi he po en ial o au oma ed analysis o e ed by compu e -
aided e al moni o ing sys em, b ough back an in e es o he
e al elec oca diog aphy (Fuchs, 2014). Howe e , aking in o
accoun he need o moni o a whole p egnancy pe iod, only
a nonin asi e app oach could be conside ed, which elies on
indi ec eco ding he FECG om elec odes loca ed on ma e nal
abdominal wall (Ungu eanu e al., 2009; Vullings e al., 2010;
Kimu a e al., 2012; Khala e al., 2013; Beha e al., 2014;
Agos inelli e al., 2015b).
Ano he impo an issue o de elopmen o e ec i e
abdominal elec oca diog aphy e e s o a g owing in e es in
high- isk p egnancy elemoni o ing a home (W obel e al.,
2015b). When using he ul asound-based e al moni o he
ansduce has o be ca e ully placed o ensu e he ul asound
beam is ocused on he e al hea . Wha ’s mo e, du ing
moni o ing session he ansduce may equi e eposi ioning
due o a change o e us posi ion. O he wise, he signal loss
occu s which may cause, in case when a woman pe o ms he
moni o ing session alone, he un ounded ea and unp edic able
eac ion. When he abdominal elec odes a e ixed on he
abdomen, he pa ien can easily e i y he signal loss, which
in ha case occu s only when one o he elec odes peels o
(Ka ounis e al., 2007; Kolomeye s and Roshche skaya, 2013;
Agos inelli e al., 2015b).
Imp o emen o he measu emen ins umen a ion, elec ode
echnology and he signal p ocessing me hods ha ha e
been no iced du ing ecen yea s, enabled o cope wi h
he p oblems connec ed wi h de elopmen o he abdominal
e al elec oca diog aphy (Ko as, 2008; Vullings e al., 2010).
The signal acqui ed om e us head is in ac “pu e” e al
elec oca diog am, whe eas he abdominal signal includes also
he ma e nal elec oca diog am (MECG) and some noise coming
mainly om muscle ac i i y (Ta alunga e al., 2009, 2014;
Ma inek e al., 2016). Thus, he c ucial s ep in ex ac ion o he
FECG om he abdominal signal is a supp ession o ma e nal
elec oca diog am while p ese ing he e al QRS complexes
(Melillo e al., 2014; Agos inelli e al., 2015a). The ene gy o
MECG is many imes highe han he ene gy o FECG, and
wha ’s mo e he equency band o bo h hese componen s pa ly
o e laps which makes simple il e ing useless (Ka ounis e al.,
2007). A numbe o di e en app oaches o MECG supp ession
and de ec ion o e al QRS complexes we e p esen ed in li e a u e
(Ungu eanu e al., 2007; Liu and Luan, 2015; Poian Da e al.,
2016). The sys em o acquisi ion o abdominal signals and
o iginal me hod o FECG ex ac ion we e p oposed by he
au ho s, and he indi ec e al elec oca diog aphy was e alua ed
in ela ion o he gold s anda d—di ec FECG app oach
(Jezewski e al., 2012). Re e ing o he esul s ob ained in ou
p e ious s udy, conce ning a compa ison o ul asound app oach
wi h he di ec FECG (Jezewski e al., 2006), we concluded ha
he abdominal e al elec oca diog aphy p o ides accu acy no
wo se han he ul asound me hod does. Howe e , in all s udies
whe e he US me hod o abdominal FECG was compa ed wi h
di ec FECG, he esul s we e ob ained only o he signals
being acqui ed du ing labo . Conside ing ha e al de elopmen ,
aking place du ing a whole p egnancy pe iod, a ec s he
cha ac e is ics o he FHR signals, we decided o ca y ou he
compa ison o he abdominal FECG and ul asound me hod
based on he signals collec ed du ing p egnancy. I is ob ious ha
such app oach excludes he di ec elec oca diog aphy om he
s udy, and causes some p oblems o compa ison me hodology
due o a lack o e e ence da a (Sa o e al., 2011; Cohen e al., 2012;
Kimu a e al., 2012). Since bo h ypes o he signals we e acqui ed
by means o wo se s o ins umen a ion, ano he impo an
p oblem has been ecognized— he FHR signals synch oniza ion,
i.e., inding he co esponding ca diac cycles. I should be no iced
ha in case o ul asound-based moni o , he FHR signal is
p o ided h ough i s ou pu only as he measu emen alues o
ins an aneous hea a e e enly spaced wi h 250 ms. On he o he
hand, he sys em o nonin asi e FECG is able o p o ide, along
wi h he e enly spaced signal, he ime e en se ies wi h du a ions
o consecu i e ca diac cycles.
In his wo k he me hodology is p oposed o compa e wo
di e en me hods o e al hea a e moni o ing. I s o iginali y
ela es o he ac ha compa ison has been ca ied ou no only
in ela ion o he co esponding ca diac cycle alues, bu also
o he clinically impo an indices desc ibing he ins an aneous
FHR a iabili y.
METHODS
The esea ch ma e ial comp ised he FHR signals acqui ed
simul aneously using he Dopple ul asound as well as he
elec oca diog aphic me hods in a g oup o 70 p egnan women.
F om a numbe o moni o ing sessions pe o med o each
pa ien , we selec ed only one eco ding acqui ed a ound 1 week
be o e deli e y, wi h a leng h o a leas 30 min (Geo gie a e al.,
2014). All he eco dings a e accompanied by in o ma ion on
e al ou come: ges a ional age a bi h, blood gas pa ame e s
pH and BE, pe cen ile o e al bi h weigh , Apga sco e,
in o ma ion abou a possible s ay in he NICU. The pa ien s we e
moni o ed by simul aneously using wo popula in ma e ni y
wa ds, sys ems o e al and ma e nal moni o ing: MONAKO
and KOMPOREL. Un o una ely, hese sys ems we e unable o
synch onize eco ded signals du ing he moni o ing session.
The ime shi be ween signals beginnings in each session could
each up o a ew minu es, whe eas in case o compa a i e
s udies he p ecise synch oniza ion is equi ed (e en on he
le el o indi idual hea bea s). Hence, he p oblem o signals
synch oniza ion has been conside ed as a signi ican challenge.
F on ie s in Physiology | www. on ie sin.o g 3May 2017 | Volume 8 | A icle 305
Jezewski e al. FHR om: aFECG s. US
As he esul o each simul aneous moni o ing session, wo
iles we e ob ained o he na i e o ma , whe e he FHR signal
is ep esen ed by he alues measu ed e enly wi h 250 ms
pe iod. Files om he MONAKO Sys em comp ise he FHR_U
signal cap u ed om he ou pu o e al moni o equipped wi h
he ul asound ansduce (Hewle -Packa d M1351). The iles
om he KOMPOREL Sys em p o ide he FHR_E signal being
de e mined on a basis o e al elec oca diog am eco ded om
he abdominal wall o he mo he . The e al elec oca diog am
is eco ded by using ou elec odes placed on he ma e nal
abdomen. The c ucial s ep in ex ac ion o he FECG om he
abdominal signal is a supp ession o ma e nal elec oca diog am
while p ese ing he e al QRS complexes (Cas illoa e al., 2013;
Ma inek e al., 2015). The p oposed me hod o he MECG
supp ession is based on sub ac ing he pa e n o ma e nal
P-QRS-T complexes and spa ial il e ing. I ensu es co ec
de e mina ion o he iducial poin s as well as he ac o s
scaling he pa e n. The algo i hm o de ec ion o he e al QRS
complexes is based on a ma ched il e ing app oach in o de o
educe he sensi i i y o in e e ences. Addi ionally, he de ec ion
is ca ied ou wi h a se o decision ules o p edic he du a ion
o he nex bea - o-bea ca diac cycle (Ma onia e al., 2006). As a
esul o he FECG analysis, he ime e en se ies is ob ained—
as he ime ma ke s when he successi e e al hea bea s we e
de ec ed— he QRS complexes, which is hen used o de e mine
he FHR_E signal as he alues wi h 250 ms pe iod (Gue e o
Ma inez e al., 2006; Almeida e al., 2013, 2014).
Signals Synch oniza ion
The p ocedu e o synch oniza ion o each pai o he FHR_U
and FHR_E signals consis ed o wo s ages. In he i s s age an
ini ial isual adjus men was suppo ed by a dedica ed p og am
o isualiza ion o he signals. This p og am as well as all he
o he s, c ea ed o he pu pose o his wo k, was de eloped
in LabView en i onmen (Na ionaI Ins umen s) (Desai e al.,
2013). A e coa se synch oniza ion o he signals he common
pa o FHR_U and FHR_E signals was sepa a ed. I elied on
mo ing he beginning and end o one signal, o indica e he
agmen o in e es acco ding o he o he signal. A his s age
he signals quali y had o be good enough o allow ecognizing he
cha ac e is ic ea u es common o bo h signals, and cons i u ing
he so called cen e ing poin s, and he common pa o he signals
had o ha e non-ze o leng h. These condi ions we e no me in
case o h ee eco dings, hence in u he p ocessing only he se
o 67 pa ien s we e included.
In he second s age he signal alida ion was conduc ed, as
well as p ecise synch oniza ion o bo h signals, a he le el o
indi idual FHR alues p o ided e e y 250 ms. The de eloped
so wa e enabled semi-au oma ed synch oniza ion. The p og am
au oma ically ound he ime shi be ween he signals o ensu e
he minimum di e ences be ween he co esponding alues,
which mos ly led o p ope synch oniza ion. A e ha , he isual
e i ica ion was ca ied ou wi h a possibili y o addi ional ime
shi co ec ion, ollowed by he inal accep ance (Figu e 1).
The so wa e o de e mina ion o he op imal ime shi
be ween he analyzed signals was using he synch oniza ion
unc ion based on he mean absolu e di e ences (MAD),
de e mined o he co esponding (applying he ime shi )
FHR_U and FHR_E alues. To imp o e he pe o mance o
synch oniza ion unc ion i was necessa y o u he educe
he in luence o andom in e e ences appea ing in he FHR
signals, as well as sudden alue changes esul ing om he
measu emen e o s o po en ial accele a ion and decele a ion
episodes. Hence, he segmen s wi h sudden changes in he FHR
signal we e excluded om he unc ion de e mina ion, i he
absolu e di e ence be ween a gi en alue and he p eceding one
was highe han 10 bpm (Spilka e al., 2012). I a gi en FHR
alue was ejec ed, he nex one was compa ed o he mean
calcula ed om he p e ious alues (including he ejec ed) in
he 240 alues window. The FHR alues we e also ejec ed om
he signal, which we e suspec ed o ep esen he ma e nal hea
a e— he de ails o he algo i hm a e p esen ed in W obel e al.
(2015b). This ype o e oneous measu emen s occu s in he
FHR_U signal, as i is ypical o he ul asound me hod and qui e
equen in he US-based e al moni o s. The abo e-men ioned
p ep ocessing is only in ended o synch onize he signals and do
no change hei in o ma ion con en .
The op imum ime shi , co esponding o he ully
synch onized signals, was ob ained o he unc ion minimum,
when applying addi ional shi in he ange om −25 o +25
FHR alues (measu ed e e y 250 ms), in espec o he signals
synch onized a e he i s s age. I a he beginning o end o
a gi en signal any in e e ence associa ed wi h he s a o end
o he moni o ing session occu ed, hey we e also emo ed in
he imming p ocess. T imming o he ull minu es in u n,
esul s om he ac ha he analysis o he ins an aneous FHR
a iabili y is always ca ied ou wi hin a 1-min signal segmen s.
Finally, as a esul o he synch oniza ion p ocedu e some signal
pai s could be sho ened by as much as 4 min. A e he second
s age o synch oniza ion he common pa o he analyzed signals
is immed o he la ges whole numbe o he minu es ( he
numbe o FHR alues was a mul iple o 240).
We assumed ha he minimum leng h o synch onized signals
(cons i u ing he pai ) subjec ed o u he analysis should be 10
min. Acco ding o ha c i e ion only one eco ding was ejec ed,
and 66 eco dings we e le .
Signal Loss Analysis
The nex p ocessing s ep consis ed o e i ying he signal pai s
in e ms o hei quali y, measu ed bo h by a size and na u e o
he signal loss episodes. Episodes o signal loss a e p elimina ily
de ec ed by he moni o ing sys ems used, and ep esen ed by
ze o alues in he FHR signal. Fo he pu pose o his wo k
also he po en ial e oneous FHR alues, as no mee ing he
adop ed c i e ia, we e ma ked as signal loss episodes, using he
dedica ed de eloped p og am. Signal segmen s we e conside ed
o be signal loss i hey did no mee he an Geijn modi ied
c i e ion ( an Geijn, 1980), p oposed in Jezewski e al. (2012).
Tha was applied o hese FHR alues, which we e conside ed
as e o s by a p ocedu e o sudden changes emo al used in
he second s age o synch oniza ion. In his case, he es ablished
h esholds a e: he absolu e di e ence be ween a gi en FHR
alue and he p eceding one g ea e han 20 bpm, and window
wid h equals o 100 alues. The segmen s, suspec ed o con ain
F on ie s in Physiology | www. on ie sin.o g 4May 2017 | Volume 8 | A icle 305
Jezewski e al. FHR om: aFECG s. US
FIGURE 1 | The sc een illus a ing he ope a ion o he p ocedu e o synch onizing he wo signals, de e mined ia he ul asound ( ed FHR_U) and
elec oca diog aphic (blue FHR_E) me hods. (A) P esen s he concep o au oma ed synch oniza ion ia minimizing he mean e o . The cu en alue o he mean
absolu e di e ence MAD pa ame e is displayed on he gauge. Au o mode had au oma ically se up he alue o he shi pa ame e o 9 samples o wa d. Howe e ,
basing on a isual analysis o he signals, ha shi was manually co ec ed using a slide — he FHR_U signal was shi ed by 2 samples backwa d in ela ion o he
FHR_E signal. (B) Shows he enla ged signal agmen om pa A, bu a e a p ocedu e o emo ing he sudden FHR changes. I allows o addi ional manual
synch oniza ion and inal alida ion o he eco ding o u he in es iga ions.
he ma e nal hea a e signal, we e indica ed as he signal loss
episodes—simila ly as in he second s age o synch oniza ion.
Finally, as a esul o he signal loss analysis, he FHR_U and
FHR_E signals we e ob ained wi h addi ional in o ma ion abou
de ec ed gaps (FHR alues equal o ze o). The signal loss le el
is de ined as a pe cen age o he du a ion o signal loss episodes
( he numbe o FHR alues equal o ze o) in ela ion o he
o al du a ion o he signal (all FHR alues). Taking in o accoun
he maximum le el o signal loss o 30% in ei he FHR_U o
FHR_E signal, i e eco dings we e emo ed. Addi ionally, in
e ms o uni o mi y o signal loss dis ibu ion in ime, ou
ques ionable eco dings wi h signal loss be ween 20 and 30%
we e isually assessed. Only one eco ding was excluded due o
he accumula ion o he signal loss (equal o 23%) in he middle
pa o he FHR_U signal. The inal esea ch ma e ial consis ed
o signal pai s om 60 moni o ing session. The o al leng h
o eco dings was equal o 1995 min. The leng h o indi idual
eco dings a ied om 11 o 64 min, wi h an a e age o 33.3 min.
Di ec Signals Compa ison
Fo he inal se o eco dings, consis ing o he FHR_U
and FHR_E signal pai s, some desc ip i e s a is ics o signal
compa ison we e calcula ed, bo h on a global basis as well
as a he le el o pa icula FHR alues. These s a is ics
include (calcula ed o indi idual eco dings): he eco ding
du a ion, signal loss le el, mean alue o he di e ences be ween
he co esponding ins an aneous FHR alues (MD), s anda d
de ia ion (SD), mean absolu e di e ence (MAD), as well as
he summa y s a is ics o he en i e esea ch ma e ial. The
di e ence be ween he pai s o co esponding ins an aneous
alues o FHR_U and FHR_E was exp essed dually: as he hea
bea s pe minu e (bpm) as well as in milliseconds. The second
ep esen a ion is ob ained by con e sion o he FHR alues
in o in e als be ween successi e hea bea s, acco ding o he
hype bolic ans o ma ion wi h he 60,000 ac o .
As equally impo an i is assumed he compa ison o he
signals in a o ma commonly used in au oma ed analysis o he
low a iabili y FHR signal componen s (Jezewski e al., 2002).
In his o ma he successi e alues o he signal a e de e mined
by a e aging 10 consecu i e o iginal FHR alues ( ime se ies
measu ed e e y 250 ms). Impo an pa in he a e aging p ocess
is he emo al o he signal loss episodes, ma ked as ze o alues.
I , while a e aging 10 o iginal FHR alues, mo e han ou alues
a e ma ked as signal loss, he esul ing alue o 2.5 s pe iod is also
conside ed as signal loss and is assigned wi h ze o alue.
Indi ec Signal Compa ison ia Clinical
Pa ame e s
As shown in he li e a u e (Jezewski e al., 2016), he di e ences
om he di ec compa ison o signals a e o en no co ela ed
F on ie s in Physiology | www. on ie sin.o g 5May 2017 | Volume 8 | A icle 305

Jezewski e al. FHR om: aFECG s. US
wi h di e ences in alues o clinically impo an pa ame e s o
quan i a i e desc ip ion o FHR signal, de e mined by he e al
moni o ing sys ems (Geo gie a e al., 2012). These pa ame e s
a e used by clinicians, in e p e ing he FHR signals in o de o
assess he e al s a e. The e o e, i was conside ed as impo an
o iden i y he impac o he FHR signal acquisi ion me hod
on he clinically signi ican pa ame e s. Fo ha pu pose, he
synch onized FHR_U and FHR_E signals we e sa ed in o he
na i e o ma iles (measu ed wi h 250 ms) and eloaded o
he a chi e o MONAKO Sys em. Thanks o an op ion o
eanalysis o he a chi al eco ds, o each o 60 eco dings
(120 FHR signals), he quan i a i e pa ame e s desc ibing he
a iabili y pa e ns de ec ed in he FHR signal we e de e mined
au oma ically. As he esul , he lis s o pa ame e s we e
ob ained o he FHR_U and FHR_E signals. They included
he pa ame e s desc ibing some pa e ns o FHR a iabili y in
he ime domain: mean alue o FHR (M_FHR), mean alue
o he FHR baseline (M_BL), numbe o de ec ed accele a ion
(ACC) and decele a ion (DEC) episodes (Geo gie a e al.,
2012; W obel e al., 2013). Addi ionally, an assessmen o
he ins an aneous FHR a iabili y was p o ided as: du a ion
and alue o he high (HE_D and HE_V) and low (LE_D
and LE_V) a iabili y episodes, a e age alue o he long-
e m a iabili y (LTV) and sho - e m a iabili y (STV) indices,
as well as he FHR oscilla ions (OSC) oge he wi h he
pe cen age o di e en oscilla ion ypes (OSC_I÷OSC_IV)
(Jezewski e al., 2016). Inconsis encies o he abo e pa ame e s
calcula ed o he co esponding FHR_U and FHR_E signals
we e es ima ed using he symme ic mean pe cen age di e ence
SMPD, whe e he di e ences be ween alues a e ela ed o
hei mean. Since he no mali y assump ion was e i ied using
he Shapi o-Wilk es , he s a is ical signi icance (using pai ed
S uden ’s - es ) o he di e ences be ween he co esponding
pa ame e s ob ained in he FHR_U and FHR_E signals was
examined.
Indi ec Signals Compa ison ia
Bea - o-Bea Va iabili y
I is gene ally belie ed ha a e y high p edic i e alue in ela ion
o he ea ly de ec ion o he e al dis ess is p o ided by he
ins an aneous FHR a iabili y pa ame e s (Cesa elli e al., 2009).
They a e de e mined om he FHR signal in a o m o he ime
e en se ies—a sequence o e en s une enly loca ed in ime,
p o iding he successi e ca diac cycles du a ion exp essed in
milliseconds.
This o ma signi ican ly di e s om ha a ailable a he
ou pu o a e al moni o — he FHR alues e enly spaced a e e y
250 ms. This measu emen pe iod has been es ablished o be
no longe han he sho es physiologically allowed hea cycle,
howe e wi h cha ac e is ic in o ma ion edundancy o low
FHR alues (e.g., he FHR alue equal o 50 bpm is ep esen ed
by ou duplica ed subsequen alues; Lee e al., 2009; Goncal es
e al., 2013).
The e o e, he FHR_U and FHR_E signals we e subjec ed
o econs uc ion o he abo e men ioned ime e en se ies
ep esen a ion. This p ocedu e elied on aking om he e enly
dis ibu ed ime se ies, he alues acco ding o he iming
signal being cons i u ed by he e al QRS complexes addi ionally
ob ained om he KOMPOREL Sys em. The esul ing signal is
a sequence o ime-o de ed e en s co esponding o subsequen
occu ences o he e al QRS complexes (o mo e p ecisely
he R-wa es). C ea ed acco ding o he desc ibed p ocedu e
he FHR_U and FHR_E signals in he o m o ime e en
se ies, p o ided he basis alues o de e mina ion o he
ins an aneous a iabili y indices. These indices, widely acclaimed
in he li e a u e (Romano e al., 2016b), quan i a i ely desc ibe
he long- and sho - e m FHR a iabili y.
In his s udy he ollowing indices we e analyzed: Haan_LTI,
Haan_STI, Yeh_II, Yeh_STI, O gan_LTV, O gan_STV,
Dal on_LTV, Dal on_STV, Zugaib_LTV, Zugaib_STV. In
o de o s anda dize he esul s he indices we e de e mined
wi hin 1-min segmen s (Kubo e al., 1987; Jezewski e al., 2006).
Each consecu i e segmen comp ised only hose ins an aneous
FHR alues which we e de e mined using he hea bea s
con ained in he gi en segmen (Cesa elli e al., 2009).
While p ocessing a gi en segmen , i pe cen age o alid
alues, ele an o a gi en index (acco ding o i s de ini ion),
was less han 20%, i was assumed ha he index alue was
unde e mined o ha minu e. Such cases a e he esul o
he signal loss episodes in he analyzed signals. In a se ies o
minu e alues calcula ed o a gi en signal, hey a e de ined
as a 1-min loss o he gi en index alue and ma ked wi h
he alue o −1. The index a e age alue o a gi en signal
is calcula ed om all 1-min alues, excluding hose ma ked
as unde e mined. Addi ionally, a non-linea pa ame e o
ins an aneous FHR a iabili y was p oposed, in a o m o
he egula i y measu e— he sample en opy index (SampEn)
(Signo ini and Magenes, 2014). I was de e mined in he
windows co e ing 300 hea e en s (FHR alues), and exp essed
in milliseconds as a measu e o pe iod. The pa ame e s o
SampEn unc ion we e se a : dim =1 and =0.1. Fo he gi en
signal, he SampEn index ep esen s he mean alue o sample
en opy de e mined in successi e windows. Inconsis encies
o indices desc ibing he ins an aneous FHR a iabili y,
de e mined o co esponding FHR_U and FHR_E signals
we e e alua ed wi h he SMPD, whe e he di e ence be ween
alues a e ela ed o hei mean. The s a is ical signi icance
(using pai ed S uden ’s - es ) o he di e ences be ween he
alues o he indices ob ained o each signal pai was also
examined.
Indi ec Signals Compa ison ia Fe al
Ou come P edic ion
Analysis o he FHR signal leads o i s classi ica ion as
co esponding o no mal o abno mal e al s a e. Since a he
ime o e al moni o ing, he e is no o he diagnos ic me hod
which could be able o con i m a co ec ness o he signal
classi ica ion, he FHR signals, being acqui ed du ing p egnancy,
a e e ospec i ely assigned o ue e al ou come (newbo n s a e)
(Chudacek e al., 2014; Romano e al., 2016a). I is jus i ied, as in
obs e ics i is assumed ha he no mal e al ou come has o be
esul o p ope e al de elopmen du ing he p egnancy pe iod.
Excluding he cases when he labo p ocess i sel caused nega i e
e ec s o he e al s a e, he same assump ion can be applied
F on ie s in Physiology | www. on ie sin.o g 6May 2017 | Volume 8 | A icle 305
Jezewski e al. FHR om: aFECG s. US
o abno mal e al ou come. I is gene ally belie ed ha his
ela ionship is main ained in case o deli e ies by cesa ean sec ion
due o he ma e nal easons. In he collec ed da abase he as
majo i y o cases we e o his ype. The 60 pa ien s ( eco dings)
we e classi ied as belonging o a no mal o abno mal g oup using
he in o ma ion on e al ou come. The abno mal s a e was se
i a leas one o he ollowing condi ions was me : Apga sco e
(a 5 min) <7, pH <7.2, BE >12, NICU s ay >24 h, o bi h
weigh pe cen ile <5%. Finally, he esea ch ma e ial included
19 eco dings wi h abno mal s a e and 41 wi h no mal s a e
assigned e ospec i ely. I is impo an ha o he majo i y o
pa ien s wi h abno mal e al ou come, he p egnancy was ended
by cesa ean sec ion due o he ma e nal indica ions (16 cases).
I sugges s ha cou se o deli e y imposed no nega i e e ec on
he newbo n. The da abase con ains six eco dings o which he
abno mal s a e was se due o h ee o mo e condi ions me .
The analysis o abili y o p edic ion o he e al ou come
was pe o med sepa a ely o he ul asound and he e al
elec oca diog aphy app oaches. Each FHR_U (FHR_E) signal
was ep esen ed by he ea u e se comp ising: 15 pa ame e s
de e mined by he MONAKO Sys em, 10 indices desc ibing he
ins an aneous FHR a iabili y and SampEn en opy measu e.
The di e ences be ween he alues o hese ea u es ob ained in
wo g oups (no mal and abno mal e al ou come) we e exp essed
by he mean pe cen age di e ence (MPD) o bo h g oups—
whe e he no mal ou come g oup was aken as e e ence. The
s a is ical signi icance o he di e ence be ween he mean alues
o each ea u e ob ained o wo g oups was assessed using
S uden ’s - es .
The indi ec compa ison o FHR_U and FHR_E signals, as
o p edic ing he e al ou come, was based on he capabili y o
classi y he FHR signals om a gi en acquisi ion me hods in o
no mal and abno mal analyzing he de e mined clinical FHR
pa ame e s.
RESULTS
Compa ison analysis conside ing wo di e en me hods o FHR
signal acquisi ion has been ca ied ou using 60 pai s o FHR_E
and FHR_U signals, which we e ob ained du ing he moni o ing
sessions o 60 pa ien s. A e a ull signals synch oniza ion
and imming, he o al leng h o eco dings was equal o
1995 min, wi h an a e age leng h o 33.3 min (SD =10.9
min). The eco dings we e cha ac e ized by low signal loss
(de ails in Table 1). Fo he US me hod he mean signal loss
was equal o 4.5%, whe eas o he FHR signals ob ained ia
FECG, he loss le el was mo e han wo imes lowe , which is
exp essed by he mean signal loss o only 1.8%. High signal
loss was obse ed in ew eco dings, especially o FHR_U.
I could be no ed ha o mo e han hal o he eco dings,
he signal loss o bo h me hods did no exceed 1%. Such low
le el o he signal loss and he su icien leng h o indi idual
eco dings enabled u he es ima ion o he inconsis ency
be ween bo h acquisi ion me hods, using he mean alues o
clinical quan i a i e pa ame e s, de e mined du ing he FHR
signal analysis.
A di ec compa ison o he FHR_E and FHR_U signals
has been based on es ima ion o he di e ences be ween he
co esponding ins an aneous FHR alues (p o ided e e y 250
ms). I enabled he me ological assessmen o he inconsis ency
be ween he wo acquisi ion me hods. Compa ison a ha s age
was pe o med o each indi idual eco ding and he summa y
o desc ip i e s a is ics o en i e esea ch ma e ial is p esen ed
in Table 1. The mean di e ence alue (MD) ob ained o all
eco dings was −0.23 bpm, which in ela ion o an a e age alue
o FHR (abou 140 bpm) gi es a ela i e e o o 0.2%. I means
ha he measu emen bias be ween he wo me hods does no
occu . When analyzing he MD alues calcula ed o pa icula
eco dings we no ed ha i did no depend on he measu ed FHR
signal alue. I is shown by he Bland-Al man plo o MD alues
agains he a e age FHR alues ( om FHR measu ed a 250 ms)
ob ained o pa icula eco dings (Figu e 2). F om he poin o
iew o bo h isual and au oma ed assessmen o he FHR signal
a iabili y, he mo e impo an seems o be he MAD, which was
equal o 1.24 bpm o all he conside ed signals. Such alue does
no a ec a isual e alua ion o he signal, since i is lowe han
he p in ing esolu ion o he FHR wa e o ms, as well as he
esolu ion o a human eye.
In he compu e -aided sys em, he au oma ed analysis aimed
a de e mina ion o clinically impo an FHR pa e ns is ca ied
ou using he FHR alues a e aged o e 2.5 s. I makes he
compa ison be ween FHR_E and FHR_U signals ep esen ed by
such a e aged alues especially impo an . A e aging p ocesses
caused a sligh dec ease o he signal loss in bo h ypes o
TABLE 1 | Desc ip i e s a is ics o he signal loss and alues o di e ences MD, MAD, de e mined o he inal se o eco dings om he
elec oca diog aphy (FECG) and he ul asound me hod (US), whe e he FHR signals we e exp essed as he o iginal 250 ms measu es, and as he alues
a e aged o e 2.5 s pe iods.
FHR 250 ms FHR 2.5 s
Signal loss (%) FECG-US (bpm) FECG-US (ms) Signal loss (%) FECG-US (bpm)
FECG US MD MAD MD MAD FECG US MD MAD
Mean 1.80 4.53 −0.23 1.24 0.71 3.83 1.35 4.29 −0.09 0.71
SD 3.01 6.52 0.38 0.46 1.25 1.62 2.59 6.42 0.40 0.43
Median 0.60 0.95 −0.19 1.12 0.59 3.36 0.35 0.75 −0.05 0.58
Min 0.00 0.10 −1.29 0.50 −4.52 1.39 0.00 0.00 −1.50 0.23
Max 14.70 26.80 1.54 2.86 5.33 9.91 12.50 26.40 1.49 2.50
F on ie s in Physiology | www. on ie sin.o g 7May 2017 | Volume 8 | A icle 305
Jezewski e al. FHR om: aFECG s. US
FIGURE 2 | Bland Al man plo showing he dependence o he mean alue o he di e ences MD (Y axis) be ween co esponding ins an aneous alues
o FHR_U and FHR_E, in ela ion o he a e age e al hea a e in he eco ding (X axis), o each o he 60 pai s o signals. The alues a e exp essed in
bea s pe minu e—bpm.
signals, o he alues: 4.3% in US, and 1.4% in FECG (Table 1).
Compa ison a his s age was pe o med o pa icula eco dings
and he summa y o desc ip i e s a is ics o en i e esea ch
ma e ial is p esen ed in Table 1. In gene al, inconsis ency
be ween FHR signals a e a e aging dec eased. In his case,
he bias be ween he wo me hods also does no occu . The
MAD alue has dec eased signi ican ly, o only 0.7 bpm, which
is below he 1 bpm le el— he minimum accu acy o he
FHR measu emen . In ela ion o he a e age alue o he
FHR signal (140 bpm), he ela i e inconsis ency was equal
o 0.5%. Conside ing hese esul s we could assume ha such
small inconsis ency should lead o he simila alues o clinical
pa ame e s p o ided by bo h me hods. Howe e , we ha e o
keep in mind ha some o hese pa ame e s a e pa icula ly
sensi i e o he FHR changes, being a esul no only o he
mean FHR di e ence, bu a he o he dis ibu ion o FHR
di e ences in ime. In con as , some o he pa ame e s o he
FHR signal a e sensi i e o empo a y high di e ences in he FHR
signals. So, in case o hese pa ame e s he di e ences be ween
bo h me hods may occu . Such o mula ed assump ions ha e
been e i ied in he nex s age o he inconsis ency analysis—
he indi ec compa ison o bo h signals. I was based on he
in e p e a ion o he di e ences be ween he pa icula FHR
signal pa ame e s, which a e p o ided by an au oma ed analysis
in he e al moni o ing sys em.
Desc ip i e s a is ics (mean alues and SD) o indi idual
pa ame e s desc ibing quan i a i ely he FHR signal, which ha e
been de e mined o signals om bo h me hods, a e p esen ed
in Table 2. Fo each pa ame e he di e ences be ween he
FHR_E and FHR_U signals, we e assessed using he SMPD. I
was jus i ied because o any o hose pa ame e s no signi ican
di e ence be ween hem was no ed. As i could be expec ed,
he low SMPD alue o 0.1% ob ained o M_FHR and M_BL
pa ame e s (being signi ican ly dependen on an a e aging
p ocess o he FHR measu emen s), was simila o he ela i e
inconsis ency epo ed in he di ec signals compa ison. In
u n, he SMPD alues calcula ed o he ollowing pa ame e s:
HE_D, LE_D, STV, ACC, DEC, OSC_IV, di e signi ican ly
om he alues epo ed o o he pa ame e s, and e en mo e
TABLE 2 | Values o clinically impo an pa ame e s o quan i a i e
desc ip ion o FHR-E and FHR-U signals, om FECG and US, ob ained in
a compu e -aided e al moni o ing sys em, oge he wi h he symme ic
mean pe cen age di e ence SMPD es ima ing he inconsis encies
be ween bo h he me hods.
Pa ame e s FHR_E FHR_U SMPD (%)
Mean SD Mean SD
M_FHR (bpm) 143.19 9.35 143.32 9.49 −0.1
M_BL (bpm) 141.70 9.55 141.79 9.60 −0.1
ACC (numbe ) 6.62 5.31 6.12 5.29 7.8
DEC (numbe ) 0.40 0.87 0.23 0.56 54.0
LTV (ms) 39.54 11.90 37.91 11.40 4.2
STV* (ms) 6.35 2.41 5.68 2.02 11.1
HE_D (min) 12.72 12.15 10.60 11.49 18.2
LE_D (min) 7.85 8.01 8.72 7.91 −10.5
HE_V (ms) 53.59 9.67 54.99 7.81 −2.6
LE_V (ms) 18.67 3.22 18.78 3.39 −0.6
OSC (ms) 13.40 3.80 12.91 3.70 3.7
OSC_I (%) 10.51 14.39 11.05 15.30 −5.0
OSC_II (%) 27.77 15.59 29.29 15.39 −5.3
OSC_III (%) 45.31 18.19 45.48 19.11 −0.4
OSC_IV (%) 8.96 9.25 7.81 9.06 13.7
*p<0.05 (pai ed - es ).
om he esul s o he di ec compa ison. I con i ms he
abo e men ioned assump ion ha some pa ame e s (e.g., STV)
a e sensi i e o dis ibu ion o he FHR di e ences in ime,
whe eas ano he ones (e.g., DEC) o a empo a y high di e ence
alue. Pa icula ly, high SMPD =54% no ed o he numbe o
ecognized decele a ions, is a esul s o wo ac o s: he di ec
di e ences be ween he signal alues and he di e ences in he
signal loss episodes. The signal loss o FHR_U is on a e age
wice highe han o FHR_E. In addi ion, he au oco ela ion
echnique, commonly used in he US me hod o de e mine he
signal pe iodici y, is o en no able o ollow he apid dec ease
o FHR signal ela ed o decele a ion, which esul s in signal loss
episodes (Figu e 3). This, in u n, causes ha he decele a ion is
F on ie s in Physiology | www. on ie sin.o g 8May 2017 | Volume 8 | A icle 305
Jezewski e al. FHR om: aFECG s. US
no ecognized, because i does no mee he es ablished c i e ia
o ampli ude and du a ion.
Indi ec compa ison o he FHR_U and FHR_E signals
was pe o med on he basis o he a iabili y indices de ined
o signal ep esen ed as ime e en se ies— he hea bea s.
Summa y o he esul s (mean alues, SD, and SMPD) o he
selec ed 11 pa ame e s desc ibing he FHR signal a iabili y is
p esen ed in Table 3. The esul s clea ly show ha he FHR_E
signal is cha ac e ized by highe a iabili y hen he FHR_U.
Inconsis encies o he long- e m a iabili y indices we e a abou
10%, whe eas o he sho - e m indices hey we e i e imes
highe , eaching abou 50%. Fo he sho - e m a iabili y we
no iced signi ican di e ence o bo h me hods.
TABLE 3 | Resul s o he FHR_E and FHR_U signal analysis, conce ning
he long- and sho - e m a iabili y, calcula ed using signal in a o m o
ime e en se ies—a sequence o e en s une enly localized in ime,
oge he wi h he SMPD alues es ima ing he inconsis encies be ween
bo h he me hods.
Index FHR_E FHR_U SMPD (%)
Mean SD Mean SD
Haan_LTI 21.89 14.69 20.62 14.59 6
Yeh_II 1022.89 1.60 2.57 1.53 12
O gan_LTV 8.18 4.46 7.35 4.35 11
Dal on_LTV 12.18 6.68 10.78 6.36 12
Zugaib_LTV 1022.22 1.25 2.01 1.21 10
Haan_STI 103# 6.29 1.94 3.10 1.02 68
Yeh_DI 103# 6.08 2.33 3.62 1.30 51
Geijn_STV* 15.27 23.70 8.98 13.51 52
Dal on_STV#1.80 0.65 1.01 0.32 56
Zugaib_STV 103# 2.77 1.02 1.76 0.58 45
SampEn+1.57 0.57 1.27 0.48 21
*p<0.03; +p<0.001; #p<0.0001 (pai ed - es ).
Wi h ega d o such la ge inconsis encies i has o be decided
which o he wo me hods may be conside ed as p o iding he
FHR a iabili y desc ip ion being close o he ue one. The
answe is no ob ious, because in his wo k no e e ence signal
was acqui ed simul aneously wi h wo analyzed me hods. Such
gold s anda d can be p o ided by p e iously men ioned he
di ec e al elec oca diog aphy, whe e he pu e FECG is acqui ed
om he e al head. In he p e ious s udies whe e he FHR signal
om ul asound me hod was compa ed wi h he e e ence one,
i has been shown ha ul asound me hod unde es ima es he
sho - e m a iabili y on a le el be ween 20 and 40% in e e ence
o di ec e al elec oca diog aphy. A simila end can be seen
in Table 3, whe e he FHR_U is compa ed wi h FHR_E ob ained
om he abdominal e al elec oca diog am.
The signal da abase has been di ided in o wo g oups: no mal
and abno mal, acco ding o he es ablished e al ou come c i e ia.
The desc ip ion o wo signal g oups is shown in Table 4.
Summa y o he FHR signal analysis esul s, comp ising 15
clinical pa ame e s de e mined o bo h g oups o e al ou come,
a e shown sepa a ely o he ul asound me hod (Table 5)
and abdominal elec oca diog aphy (Table 5). In addi ion o
he mean alue and s anda d de ia ion, he mean pe cen age
di e ence MPD was calcula ed, assuming he alues ob ained
in no mal e al ou come g oup as he e e ence. Apa om
assessing he s a is ical signi icance o he di e ence be ween he
no mal and abno mal g oups, also he endency o changes was
s udied o he pa icula pa ame e s be ween hese g oups. I
was ca ied ou o check whe he he obse ed endency would
be consis en wi h he clinical in e p e a ion o hose pa ame e s.
Conside ing he numbe o ACC pa e ns, a signi ican di e ence
be ween he abno mal and no mal g oups was no ed, and he
endency was consis en . Bu o he numbe o DEC pa e ns no
signi ican di e ence was obse ed. The MPD ook he opposi e
alues: nega i e alue o ul asound (−48%) and posi i e o
elec oca diog aphy (38%). Thus, a clinical in e p e a ion o
decele a ions as a sign o e al dis ess has been con i med
FIGURE 3 | An example o a 12-min agmen o signal pai — esul o an indi ec compa ison o e alua e he impac o he FHR signal acquisi ion
me hod on clinically ele an pa ame e s, de e mined by a compu e -aided e al moni o ing sys em. The au oco ela ion echnique, commonly used in he
US me hod o de e mine he signal pe iodici y, is o en no able o ollow he apid dec ease o FHR_U signal ela ed o decele a ion, which esul s in signal loss
episodes. This, in u n, causes ha he decele a ion is no ecognized, because i does no mee he es ablished c i e ia o ampli ude and du a ion. G aphic ma ke s o
he analysis esul s illus a e he signal loss (abo e he cu e), he es ima ed FHR baseline (line i ed on FHR cu e) and de ec ed decele a ion episodes (ho izon al
ba s unde he cu e).
F on ie s in Physiology | www. on ie sin.o g 9May 2017 | Volume 8 | A icle 305