37
Leka a echnika – Clinician and Technology 2024, ol. 54(2), pp. 37–42, DOI: 10.14311/CTJ.2024.2.01
ISSN 0301-5491 (P in ), ISSN 2336-5552 (Online)
ORIGINAL RESEARCH
ENHANCING CLINICAL EFFICIENCY: AUTONOMOUS
DETERMINATION OF CARDIAC EFFECTIVE REFRACTORY PERIOD
USING ECG SIGNALS
Richa d Redina1, 2, Ma ina Filipenska1
1Depa men o Biomedical Enginee ing, Facul y o Elec ical Enginee ing and Communica ions,
B no Uni e si y o Technology, B no, Czech Republic
2In e na ional Clinical Resea ch Cen e, S . Anne’s Hospi al, B no, Czech Republic
Abs ac
Elec ophysiological p ocedu es o managing a hy hmias emain ime-consuming, causing pa ien s o wai o c i ical
in e en ions. To add ess his p essing need, we in oduce a e olu iona y algo i hm ha au onomously calcula es he
e ec i e e ac o y pe iod (ERP) o ca diac issue om elec oca diog am (ECG). The algo i hm p incipally comp ises
signal il e ing echniques and he de ec ion o local ex ema wi hin he signal wa e o m. This algo i hm unde wen
igo ous assessmen using an in-house da abase o ECG signals acqui ed om en pa ien s who unde wen
elec ophysiological examina ions. Fundamen al digi al signal p ocessing me hods, such as linea il e ing and
h esholding, we e employed in he de e mina ion o ERP. The algo i hm yielded esul s cong uen wi h he ERP alues
es ablished by elec ophysiologis s in nine ou o en cases, wi h a s anda d de ia ion o 18.97 milliseconds. By being
accu a e and easy o in eg a e., his algo i hm holds p omise o eal- ime deploymen in clinical se ings, whe e i could
po en ially s eamline and au oma e s imula ion p o ocols, he eby expedi ing he examina ion p ocess.
Keywo ds
Ca diac Elec ophysiology, E ec i e Re ac o y Pe iod, A hy hmology, S1–S2 S imula ion, In aca diac Elec og am
In oduc ion
In ou con empo a y landscape, he p e alence o
ca diac ailmen s is s eadily moun ing, emphasizing he
c i ical need o imp o ed ea men modali ies [1].
Wi hin his con ex , elec ophysiology eme ges as
a ealm ipe o inno a i e b eak h oughs in ca diac
ca e.
Elec ophysiology is a specialized ield dedica ed o
he in es iga ion o elec ical phenomena in biological
sys ems, wi h a p ima y ocus on he human body. This
discipline aiming on he p ecise measu emen o
elec ical signals o igina ing wi hin cells, issues, and
o gans, all in he pu sui o deepe insigh s in o
pa hological condi ions [2].
The co ne s one o elec ophysiology's app oach o
managing ca diac a hy hmias is he elec ophysio-
logical examina ion [3]. This p ocedu e is pi o al in i s
dual ole o elucida ing issue cha ac e is ics unde
sc u iny, iden i ying abe an conduc ion pa hways
be ween a ia and en icles, and pinpoin ing he o igins
o ca diac a hy hmias. Mo eo e , i encompasses
he apeu ic dimensions, in ol ing he isola ion o
a ec ed myoca dial segmen s h ough he applica ion
o adio equency ene gy.
Ou esea ch is chie ly ocused on s eamlining he
diagnos ic phase o his p ocedu e, pa icula ly in
au oma ing he de e mina ion o he e ec i e e ac o y
pe iod (ERP) o he conduc ion sys em. Ou ul ima e
goal is o expedi e he en i e examina ion p ocess, which
is inhe en ly ime-consuming, h ough he au oma ion o
his c i ical s ep. Addi ionally, we p o ide a p o icien
and unbiased diagnos ic ool o he a ending physician.
This endea ou ma ks a signi ican s ide owa d
enhancing he e iciency and p ecision o ca diac
examina ions and ea men s.
38
Leka a echnika – Clinician and Technology 2024, ol. 54(2), pp. 37–42, DOI: 10.14311/CTJ.2024.2.01
ISSN 0301-5491 (P in ), ISSN 2336-5552 (Online)
ORIGINAL RESEARCH
Me hods
Da ase
An in e nal da abase o signals ob ained du ing
elec ophysiological examina ions conduc ed a he
Uni e si y Hospi al B no was u ilized o he de elop-
men o he algo i hm o au oma ed ERP measu emen .
All pa icipan s p o ided w i en in o med consen .
A o al o en signals we e andomly selec ed om
a ious pa ien s unde going ERP measu emen s o he
a io en icula (AV) node.
The signal du a ions anged om 57 o 208 seconds.
The S . Jude Wo kMa e 4.2 EP, S . Jude Medical, USA
sys em was employed o signal acquisi ion, wi h
a sampling equency o 2000 Hz and a ol age
esolu ion o 78 nV/LSB. The acquisi ion sys em is
equipped wi h a buil -in band-s op il e o supp ess he
50 Hz equency and a high-pass il e wi h a cu -o
equency o 0.1 Hz o elimina e low- equency noise.
Du ing he p ocedu e, pa ien s we e subjec ed o
con en ional wel e-lead su ace ECG measu emen s
alongside i e in aca diac elec og am leads. These
in aca diac elec og ams we e eco ded using en-pole
ca he e s posi ioned in he co ona y sinus (CS).
S1–S2 s imula ion
A c ucial aspec o he eco ding p ocess in ol ed S1–
S2 s imula ion, a widely employed p o ocol o
assessing he elec ophysiological cha ac e is ics o
he issue unde in es iga ion, pa icula ly in ERP
de e mining [4]. This p o ocol en ails epe i i e
s imula ion o he issue wi h an S1 in e al, succeeded
by one o mo e s imuli a an S2 in e al, ypically
sho e .
The p ecise numbe o s imuli and he du a ion o he
S1 phase can exhibi a iabili y. By inc emen ally
dec easing he leng h o he S2 in e al, a h eshold is
eached whe e he examined issue can no longe
espond o he s imulus. The ERP o he issue is hen
de e mined as he sho es du a ion o he S2 in e al o
which he issue emains esponsi e. A isual
ep esen a ion o his p o ocol is p esen ed in Fig. 1.
When assessing he ERP o he AV node, s imula ion
is conduc ed a he a ial le el (CS leads), wi h esponse
moni o ing aking place a he en icula le el (su ace
ECG).
Au oma ic ERP measu emen
The p oposed me hodology o au oma ed ERP
de ec ion is elegan ly s aigh o wa d ye highly
e icien , encompassing a se ies o p e-p ocessing and
analy ical s eps applied o bo h in aca diac and su ace
eco dings (see Fig. 2).
Time (s)
Fig. 1: S1–S2 s imula ion. Numbe o S1 s imuli
(T = 500 ms) a e ollowed by sho e S2 s imulus
T = 300 ms (a) o T = 290 ms (b). The las one
ansduced is a s imulus wi h a pe iod o 300 ms; sho e
S2 was no ansduced (black a ows).
39
Leka a echnika – Clinician and Technology 2024, ol. 54(2), pp. 37–42, DOI: 10.14311/CTJ.2024.2.01
ISSN 0301-5491 (P in ), ISSN 2336-5552 (Online)
ORIGINAL RESEARCH
Fig. 2: Block scheme o he algo i hm o au oma ed
ERP de ec ion.
Ini ially, a lead showcasing exclusi ely s imula ion
spikes, cha ac e ized by he highes ampli udes among
all eco ded signals, is singled ou . Wi hin his lead,
s imula ion spikes a e u he pinpoin ed as de ia ions
mee ing speci ic c i e ia: hei peak magni ude mus
be a leas 75% o he signal's maximum alue,
wi h adjacen spikes sepa a ed by a minimum o
150 milliseconds. In he in e als be ween indi idual
s imula ions, he sea ch commences o subsequen S2
pe iods sho e han he p eceding S1 pe iod.
Subsequen ly, he analysis shi s o su ace lead V6,
which unde goes il a ion. I is subjec ed o bo h a low-
pass il e wi h a cu -o equency o 45 Hz and a high-
pass il e wi h a cu -o equency o 15 Hz, u ilizing
a ini e impulse esponse (FIR) il e o de o 35. This
p ocess accen ua es he QRS complexes while
supp essing unwan ed componen s [5].
Following his, segmen s las ing 500 milliseconds a e
selec ed a e he al eady iden i ied S2 pe iods,
beginning 25 milliseconds a e he pe iod's ends. These
segmen s a e p esumed o con ain he hea 's esponse o
s imula ion, mani es ing as a QRS complex in he
su ace ECG. The e o e, QRS complex de ec ion is
pe o med in each selec ed segmen by compa ing he
maximum de ia ion o he segmen wi h an empi ically
se h eshold o 2 µV. A sub h eshold maximum
indica es he absence o a QRS complex in he segmen
ollowing s imula ion, signi ying he absence o issue
esponse o such a b ie s imula ion pe iod. The las S2
pe iod ea u ing a p esen QRS complex is iden i ied as
he ERP and epo ed o u he u iliza ion.
Model obus ness
One o he c ucial a ibu es o any signal p ocessing
algo i hm is i s obus ness agains noise ha may be
p esen in he da a. The e can be se e al ypes o such
noise, and o he pu poses o his wo k, wo we e
selec ed, he occu ence o which is nea ly una oidable.
The i s one is andom noise, speci ically Gaussian
noise. This ype o noise is cha ac e ized by
a p obabili y densi y unc ion equal o he no mal
dis ibu ion. The second ype o noise is powe line
in e e ence (PLI). The p esence o his ype o noise is
almos ine i able unde common condi ions due o he
ubiqui ous elec ical g id. An ad an age o his noise is
i s na ow spec al cha ac e is ic, p ima ily limi ed o
he equency o 50 Hz (o 60 Hz in he US).
Bo h ypes o noise we e a i icially in oduced in o
he signals, and a ious le els o signal- o-noise a io
(SNR) we e es ed o e alua e he algo i hm's
pe o mance unde challenging condi ions. The
obus ness o he model was es ed a a o al o i e SNR
le els (20, 15, 10, 5, and 0 dB).
Resul s
ERP es ima ion
The p esen ed algo i hm was applied o en eco ds
om elec ophysiological examina ions, and he
algo i hm's ou pu s a e summa ized in Table 1. In nine
ou o en cases, he ERP o he AV node was co ec ly
iden i ied. In one case, he ERP o he AV node was no
co ec ly iden i ied due o he p esence o a pa hological
conduc ion pa hway wi h a lowe ERP han ha o he
AV node. Consequen ly, he s imula ion p opaga ed o
he en icles. Howe e , he p opaga ed s imulus was
delayed and exhibi ed al e ed mo phology. This case is
illus a ed in Fig. 3.
40
Leka a echnika – Clinician and Technology 2024, ol. 54(2), pp. 37–42, DOI: 10.14311/CTJ.2024.2.01
ISSN 0301-5491 (P in ), ISSN 2336-5552 (Online)
ORIGINAL RESEARCH
Time (s)
Fig. 3: No iceable change in he mo phology o he QRS
complex ma king he eaching o he ERP accesso y
pa hway (a). Exceeding he ERP o he accesso y
pa hway esul s in he disappea ance o he
cha ac e is ic del a wa e o he second case (b).
Robus ness o a i icial noise
As demons a ed in he s udy [6], he commonly
measu ed SNR le els in signals ypically ho e a ound
10 dB, a le el ha he model should be able o handle.
In he con ex o ou expe imen , we independen ly
co up ed he su ace ECG signal wi h bo h ypes o
noise. The algo i hm's pe o mance on he co up ed
da a is summa ized in Tables 2 and 3. The bolded alues
in he ables highligh he di e ence om he anno a ion
p o ided by he physician.
Table 2: ERP alues (ms) in indi idual pa ien s unde
di e en Gaussian noise le el (dB).
The las ow
ep esen s he s anda d de ia ion (ms) o he whole
se .
SNR
Pa ien
20
15
10
5
0
1
270
270
270
270
270
2
250
250
250
250
250
3
300
300
300
300
300
4
260
260
260
260
250
5
320
320
320
320
380
6
220
220
220
330
220
7
340
340
330
340
330
8
320
320
320
320
320
9
400
400
400
400
400
10
310
310
310
300
340
σ
18.97
18.97
18.97
19.24
34.79
Al hough inc easing a ia ions wi h inc easing noise
le el a e cap u ed in he ables, a wo-way ANOVA es
did no con i m a signi ican e ec o in ensi y le el o
noise ype (p > 0.05). Thus, i can be said ha he
p oposed me hod does no show s a is ically signi ican
de ia ions in ei he ca ego y.
Table 1: Resul ing ERP alues (ms) measu ed du ing
he p ocedu e by he elec ophysiologis and
ob ained using he algo i hm. The las ow ep esen s
he s anda d de ia ion (ms) o he whole se .
Pa ien
Physician
Algo i hm
1
270
270
2
310
250
3
300
300
4
260
260
5
320
320
6
220
220
7
340
340
8
320
320
9
400
400
10
310
310
σ
-
18.97
41
Leka a echnika – Clinician and Technology 2024, ol. 54(2), pp. 37–42, DOI: 10.14311/CTJ.2024.2.01
ISSN 0301-5491 (P in ), ISSN 2336-5552 (Online)
ORIGINAL RESEARCH
Table 3: ERP alues (ms) in indi idual pa ien s
unde di e en powe line in e e ence noise (dB).
The las ow ep esen s he s anda d de ia ion (ms)
o he whole se .
SNR
20
15
10
5
0
Pa ien
1
270
270
270
270
270
2
250
250
250
250
250
3
300
300
300
300
300
4
260
250
250
250
250
5
320
320
320
320
320
6
220
220
220
220
220
7
330
330
330
330
330
8
320
320
320
320
320
9
400
390
390
390
390
10
300
300
300
300
300
σ
18.97
19.49
20.00
20.00
20.00
Discussion
The algo i hm p esen ed by us se es as a undamen al
building block o he au oma ed assessmen o ERP
du ing elec ophysiological p ocedu es. In ou pape , we
show ha ERP measu emen s can be la gely au oma ed
and hus sa e ime du ing elec ophysiological
p ocedu es. Despi e p omising esul s, his is only
a p oo o concep .
One o he limi a ions o he algo i hm could
undoub edly be he ela i ely small da ase on which i
was es ed. Un o una ely, a small da ase may no
adequa ely cap u e he wide in e indi idual a iabili y
p esen among di e en pa ien s. Addi ionally, he
algo i hm does no accoun o he possible p esence
o accesso y conduc ion pa hways, which may be
pa icula ly ele an in younge pa ien s.
As e iden om he ables, he algo i hm's
pe o mance de e io a es wi h inc easing noise le els.
To compa e he esul s o he indi idual ypes o noise,
we can use he s anda d de ia ion o he e o om he
o iginal anno a ion. A clea compa ison eme ges,
showing ha he algo i hm handles PLI noise be e .
This is likely due o he ac ha he signal passes
h ough a low-pass il e wi h a cu -o equency o
45 Hz, which should e ec i ely il e ou he unwan ed
in luence o PLI.
Con e sely, he wideband Gaussian noise is emo ed
less e ec i ely, and he s anda d de ia ion alue almos
doubles wi h inc easing SNR. To achie e be e noise
emo al, i would be necessa y o add an addi ional il e
o educe he cu -o equency o he men ioned high-
pass il e .
The esul s ob ained by ou me hod no only
co espond o he esul s ob ained manually by an
elec ophysiologis du ing he p ocedu e, bu also align
wi h he alues ob ained expe imen ally in he s udy [7].
Despi e he men ioned limi a ions, he e is signi ican
po en ial o es ing he algo i hm on addi ional eco ds
om di e en pa ien s. Ano he a enue o expanding
he algo i hm is conside ing he examina ion o o he
issues, hus in ol ing he analysis o a ious lead
combina ions. An ancilla y p oduc o he algo i hm is
he iden i ica ion o he s imula ed hy hm, which can
p o ide aluable in o ma ion o aining deep lea ning
sys ems.
Since one o he limi a ions men ioned is he ela i ely
small da ase used o es ing, pe o ming c oss-
alida ion on la ge da ase s om di e se pa ien
popula ions can p o ide a mo e comp ehensi e
assessmen o he algo i hm's pe o mance and i s abili y
o gene alize ac oss di e en pa ien coho s.
Gi en he algo i hm's inabili y o accoun o
accesso y conduc ion pa hways, conduc ing sensi i i y
analyses o simula ions o assess he impac o hese
pa hways on algo i hm pe o mance could be aluable.
This could in ol e in oducing simula ed accesso y
pa hways in o he da a and e alua ing how he algo i hm
esponds.
Conclusion
In ou pape , an algo i hm o he au oma ic de ec ion
o AV node ERP was in oduced. The algo i hm
demons a ed success in nine ou o en cases,
es ablishing i s eliabili y. Fu he de elopmen o he
algo i hm could po en ially lead o ully au oma ed eal-
ime measu emen o he ERP in a ious pa s o he
conduc ion sys em du ing examina ions.
The use o ou me hod can educe he ime o S1–S2
s imula ion, which can ake se e al minu es. The
au oma ic assessmen o he ca diac issue esponse can
ee he hands o he physician, who can ocus on o he
ac i i ies du ing he examina ion—e.g. inse ing
he apeu ic ca he e s.
Pa ial esul s om his wo k we e p e iously
published a he T ends in Biomedical Enginee ing 2023
con e ence [8].
Acknowledgemen
B no Ph.D. Talen Schola ship Holde unded by he
B no Ci y Municipali y.
Du ing he p epa a ion o his wo k he au ho s used
OpenAI's Cha GPT and Google Ba d o p e-co ec ing
English syn ax and g amma . A e using his
ool/se ice, he au ho s e iewed and edi ed he con en
as needed and ake ull esponsibili y o he con en o
he publica ion.

42
Leka a echnika – Clinician and Technology 2024, ol. 54(2), pp. 37–42, DOI: 10.14311/CTJ.2024.2.01
ISSN 0301-5491 (P in ), ISSN 2336-5552 (Online)
ORIGINAL RESEARCH
Re e ences
[1] Vadugana han M, Mensah GA, Tu co JV, Fus e V, Ro h GA.
The Global Bu den o Ca dio ascula Diseases and Risk:
A Compass o Fu u e Heal h. Jou nal o he Ame ican College
o Ca diology. 2022 No ;80(25):2361–71.
DOI: 10.1016/j.jacc.2022.11.005
[2] Ka i sis DG, Bo iani G, Cosio FG, Hind icks G, Jaïs P,
Josephson ME, e al. Eu opean Hea Rhy hm Associa ion
(EHRA) consensus documen on he managemen o
sup a en icula a hy hmias, endo sed by Hea Rhy hm Socie y
(HRS), Asia-Paci ic Hea Rhy hm Socie y (APHRS), and
Sociedad La inoame icana de Es imulación Ca diaca
y Elec o isiologia (SOLAECE). EP Eu opace.
2016 No 17;19(3):465–511. DOI: 10.1093/eu opace/euw301
[3] Ma in E, Alan B, Ma in F. Základy s deční elek o yziologie
a ka é o ých ablací. G ada Publishing a.s., 2012. 264 p.
[4] No hs ein M, Luik A, Jadidi A, Sánchez J, Unge LA, Wül e s
EM, e al. CVAR-Seg: An Au oma ed Signal Segmen a ion
Pipeline o Conduc ion Veloci y and Ampli ude Res i u ion.
F on ie s in Physiology. 2021 May 24;12:673047.
DOI: 10.3389/ phys.2021.673047
[5] Thako NV, Webs e JG, Tompkins WJ. Es ima ion o QRS
Complex Powe Spec a o Design o a QRS Fil e . IEEE
T ansac ions on Biomedical Enginee ing. 1984 No ;BME-
31(11):702–6. DOI: 10.1109/TBME.1984.325393
[6] Blanco-Velasco M, Weng B, Ba ne KE. ECG signal denoising
and baseline wande co ec ion based on he empi ical mode
decomposi ion. Compu e s in Biology and Medicine.
2008 Jan;38(1):1–13. DOI: 10.1016/j.compbiomed.2007.06.003
[7] Co ino VD, Sandbe g F, Pla ono PG, Maina di LT, Ulimoen
SR, Enge S, e al. Non-in asi e e alua ion o he e ec o
me op olol on he a io en icula node du ing pe manen a ial
ib illa ion. Eu opace. 2014 Oc 31;16(suppl_4):i 129–34.
DOI: 10.1093/eu opace/euu246
[8] Ředina R, Ronzhina M. Es ima ion o myoca dial conduc ion
eloci y using a co ona y sinus ca he e . In: No ak V, edi o .
P oceedings II o he 29 h Con e ence STUDENT EEICT 2023;
2023 Ap 25. B no: B no Uni e si y o Technology, Facul y o
Elec ical Enginee ing and Communica ion; c2023.
p. 156–60. DOI: 10.13164/eeic .2023.156
Richa d Ředina
Depa men o Biomedical Enginee ing
Facul y o Elec ical Enginee ing and Communica ion
B no Uni e si y o Technology
An onínská 2, CZ-601 00, B no
E-mail: 195715@ u .cz
Phone: +420 541 146 676