E ik Alonso, Jesús Ruiz, Elisabe e A amendi, Digna González-O e o, So ía Ruiz de Gauna, Unai
Ayala, James K Russell, Mohamud Daya. Reliabili y and accu acy o he ho acic impedance
signal o measu ing ca diopulmona y esusci a ion quali y me ics. Resusci a ion, Volume 88,
2015, Pages 28-34, ISSN 0300-9572, h ps://doi.o g/10.1016/j. esusci a ion.2014.11.027.
(h ps://www.sciencedi ec .com/science/a icle/pii/S0300957214008818)
Abs ac
Aim: To de e mine he accu acy and eliabili y o he ho acic impedance (TI) signal o assess
ca diopulmona y esusci a ion (CPR) quali y me ics.
Me hods: A da ase o 63 ou -o -hospi al ca diac a es episodes con aining he comp ession
dep h (CD), capnog aphy and TI signals was used. We de eloped a ches comp ession (CC) and
en ila ion de ec o based on he TI signal. TI shows luc ua ions due o CCs and en ila ions. A
decision algo i hm classi ied he local maxima as CCs o en ila ions. Se en CPR quali y me ics
we e compu ed: mean CC- a e, ac ion o minu es wi h inadequa e CC- a e, ches
comp ession ac ion, mean en ila ion a e, ac ion o minu es wi h hype en ila ion,
ins an aneous CC- a e and ins an aneous en ila ion a e. The CD and capnog aphy signals
we e accep ed as he gold s anda d o CC and en ila ion de ec ion espec i ely. The accu acy
o he de ec o was e alua ed in e ms o sensi i i y and posi i e p edic i e alue (PPV).
Dis ibu ions o each me ic compu ed om he TI and om he gold s anda d we e
calcula ed and es ed o no mali y using one sample Kolmogo o -Smi no es . Fo no mal
and no no mal dis ibu ions, wo sample - es and Mann-Whi ney U es espec i ely we e
applied o es o equal means and medians espec i ely. Bland-Al man plo s we e
ep esen ed o each me ic o analyze he le el o ag eemen be ween alues ob ained om
he TI and gold s anda d.
Resul s: The CC/ en ila ion de ec o had a median sensi i i y/PPV o 97.2%/97.7% o CCs and
92.2%/81.0% o en ila ions espec i ely. Dis ibu ions o all he me ics showed equal
means o medians, and ag eemen s >95% be ween me ics and gold s anda d was achie ed
o mos o he episodes in he es se , excep o he ins an aneous en ila ion a e.
Conclusion: Wi h ou da a, he TI can be eliably used o measu e all he CPR quali y me ics
p oposed in his s udy, excep o he ins an aneous en ila ion a e.
Keywo ds: Au oma ed ex e nal de ib illa o ; Ca diopulmona y esusci a ion quali y; Ches
comp ession; Tho acic impedance; Ven ila ions.
Reliabili y and accu acy o he ho acic impedance signal o measu ing
ca diopulmona y esusci a ion quali y me ics
E ik Alonso∗,a, Jes´us Ruiza, Elisabe e A amendia, Digna Gonz´alez-O e oa, So ´ıa Ruiz de Gaunaa,
Unai Ayalaa, James K. Russellb, Mohamud Dayac
A ilia ion and add esses:
aCommunica ions Enginee ing Depa men .
Uni e si y o he Basque Coun y UPV/EHU.
Alameda U quijo S/N
48013 Bilbao, Spain
bPhilips Heal hca e.
Bo hell, WA 98021, Uni ed S a es
cDepa men o Eme gency Medicine.
O egon Heal h & Science Uni e si y.
97239-3098 Po land, OR, Uni ed S a es
Co esponding au ho :
∗
E ik Alonso
email: e ik [email protected]
Tel. : +34946017384
Fax. : +34946014259
Wo d coun s:
Abs ac : 258 wo ds
Pape : 3085 wo ds
Abs ac
Aim: To de e mine he accu acy and eliabili y o he ho acic impedance (TI) signal o assess
ca diopulmona y esusci a ion (CPR) quali y me ics.
Me hods: A da ase o 63 ou -o -hospi al ca diac a es episodes con aining he comp ession
dep h (CD), capnog aphy and TI signals was used. We de eloped a ches comp ession (CC) and
en ila ion de ec o based on he TI signal. TI shows luc ua ions due o CCs and en ila ions. A
decision algo i hm classi ied he local maxima as CCs o en ila ions. Se en CPR quali y me ics
we e compu ed: mean CC- a e, ac ion o minu es wi h inadequa e CC- a e, ches comp ession
ac ion, mean en ila ion a e, ac ion o minu es wi h hype en ila ion, ins an aneous CC- a e
and ins an aneous en ila ion a e. The CD and capnog aphy signals we e accep ed as he gold
s anda d o CC and en ila ion de ec ion espec i ely. The accu acy o he de ec o was e alua ed
in e ms o sensi i i y and posi i e p edic i e alue (PPV). The eliabili y o he TI o measu e
CPR quali y me ics was analyzed by compa ing s a is ically he me ics ob ained om he TI wi h
hose ob ained om he gold s anda d. The accu acy o each me ic was analyzed indi idually o
each episode.
Resul s: The CC/ en ila ion de ec o had a median sensi i i y/PPV o 97.2%/97.7% o CCs and
92.2%/81.0% o en ila ions espec i ely. Dis ibu ions o all he me ics showed equal means o
medians, and ag eemen s >95% be ween me ics and gold s anda d was achie ed o mos o he
episodes in he es se , excep o he ins an aneous en ila ion a e.
Conclusion: Wi h ou da a, he TI can be eliably used o measu e all he CPR quali y me ics
p oposed in his s udy, excep o he ins an aneous en ila ion a e.
Keywo ds
Ca diopulmona y Resusci a ion Quali y, Tho acic Impedance, Ches Comp ession, Ven ila ions,
Au oma ed Ex e nal De ib illa o
1. INTRODUCTION1
Deli e y o high quali y ca diopulmona y esusci a ion (CPR) is a key componen o cu en 2
esusci a ion guidelines.1,2 Minimally in e up ed ches comp essions (CCs), wi h a a e o a leas 3
100 min−1and a leas 5 cm dep h, a e ecommended along wi h allowance o ull ches ecoil on4
comple ion o each comp ession. Guidelines also ecommend oxygena ion o be p o ided wi h a5
en ila ion a e o 8-10 min−1in in uba ed pa ien s. Se e al s udies ha e e ealed ha low quali y6
CCs, pauses in CCs, and hype en ila ion a e associa ed wi h poo e ou comes in bo h animals3–9
7
and humans.10–14 Ne e heless, subop imal quali y o CPR is common in bo h in-hospi al and8
ou -o -hospi al ca diac a es (OHCA).7,9,12–18
9
P ocedu es such as episode deb ie ing o he inco po a ion o eedback sys ems in o de ib illa o s10
ha e been de eloped o imp o e CPR quali y. A s anda dized e iew o esusci a ion episodes11
pe mi s epo ing o escue pe o mance in e ms o CPR quali y me ics ela ed o CCs and12
en ila ions. E o s ha e been made o s eamline hese epo s by de ining he CPR quali y13
me ics ha should be epo ed.19 Sys ems o eal ime eedback on CPR ha e p o en e ec i e14
o imp o e me ics.20 The eedback on oo slow CCs o hype en ila ion can help guide escue s15
owa ds me ics ecommended by guidelines.16,21–24
16
In he con ex o ad anced li e suppo (ALS), he moni o /de ib illa o s may be equipped wi h17
accele a ion/ o ce senso s as well as capnog aphy and pulse oxime y modules. These may be used18
o assis he escue on he CPR quali y. By con as , in he con ex o basic li e suppo (BLS),19
some au oma ed ex e nal de ib illa o s (AEDs) also o e accele a ion/ o ce senso s o eedback,20
bu hese a e ela i ely expensi e accesso ies o widesp ead use. O en he only pa ien in e ace21
is de ib illa ion pads, and only he elec oca diog am (ECG) and TI signals a e a ailable.22
E e y CC causes a luc ua ion isible in he TI signal. The TI signal can be used o iden i y23
CCs16,25,26 and en ila ions.16,27–30 Algo i hms o CC de ec ion using he TI signal a e in eg a ed24
in o comme cial so wa e o episode e iewing, and algo i hms o en ila ion de ec ion ha e been25
ecen ly p oposed using he TI.27,28 Thei global pe o mance is p o ided in e ms o sensi i i y26
and posi i e p edic i e alue (PPV).25 Ne e heless, de ailed analysis o he u ili y o he TI signal27
o CC/ en ila ion me ics is needed.28
The pu pose o his s udy was o analyze e ospec i ely OHCA episodes o de e mine he29
eliabili y and accu acy o using he TI o e alua e se en quali y me ics o he CCs and he30
en ila ions. An e icien de ec o o CCs and en ila ions was de eloped, and he eliabili y and31
accu acy o each o he se en CPR me ics was e alua ed globally and o each pa ien .32
2. MATERIALS AND METHODS33
2.1. Da a ma e ials34
The da ase used in his s udy was a subse o a la ge ou -o -hospi al ca diac a es (OHCA)35
egis y con aining 623 episodes main ained by he Tuala in Valley Fi e & Rescue agency (Tiga d,36
O egon, USA). The episodes, one pe pa ien , we e collec ed using he Philips Hea S a MRx37
moni o /de ib illa o be ween 2006 and 2009. F om he o iginal da abase only 199 episodes had38
he comp ession dep h signal (CD), he TI signal and he capnog am. Those episodes whe e39
he h ee signals we e concu en o a leas 20 min we e selec ed. A o al o 63 episodes wi h40
mean (SD) du a ion 41 (11) min comp ised he da ase . The CD signal (sampling a e 250 Hz)41
was compu ed om he o ce and he accele a ion eco ded h ough a CPR assis pad. The TI42
signal was eco ded h ough he de ib illa ion pads by applying a sinusoidal exci a ion cu en 43
(32 kHz, 3 mA peak- o-peak) wi h a esolu ion o 0.74 mΩ pe leas signi ican bi , a bandwid h44
o 0 −80 Hz and a sampling a e o 200 Hz. The capnog am was acqui ed using Mic os eam45
(sides eam acquisi ion) wi h a equency a e o 40 Hz and a esolu ion o 0.004 mmHg pe bi .46
Fo he CC-de ec ion he CD was used as gold s anda d. The ins an s o CCs we e au oma ically47
ma ked applying a ixed h eshold o -15 mm and manually e iewed (panel a in Fig. 1). Fo 48
en ila ion de ec ion, he capnog am was used as gold s anda d and en ila ions we e anno a ed49
independen ly by h ee expe ienced biomedical enginee s elying on isual inspec ion (panel a in50
Fig. 1). In e als whe e capnog aphy was disconnec ed (0.90% o he ime) o unin e p e able51
because he en ila ion pa e n was no clea ly ecognizable (11.65% o he ime), we e excluded52
om he analysis. A o al o 2575 minu es we e analyzed, which included 110286 CCs and 1758653
en ila ions. Episodes we e andomly alloca ed o aining and es se s, 32 and 31 episodes54
espec i ely.55
2.2. Simul aneous CC/ en ila ion de ec o 56
An algo i hm ha simul aneously de ec s CCs and en ila ions was de eloped based on he57
TI signal. The TI signal, z[n], whe e nis he ime sample numbe , was digi ally p ep ocessed.58
Local maxima we e iden i ied, and wa e o m ea u es cha ac e izing he signal in hei icini y59
we e ex ac ed. A decision algo i hm based on he wa e o m ea u es de e mined whe he o no 60
each local maximum was CC o en ila ion (Fig. 2).61
Signal p ep ocessing62
The z[n] signal was p ep ocessed as shown in Fig. 2. Fo CC de ec ion, a low-pass il e wi h63
a cu -o equency o c1= 1.8 Hz was applied in o de o emo e high equency noise and o64
e ain luc ua ions due o CCs and en ila ions. The esul ing signal is deno ed as zc[n]. Fo 65
en ila ion de ec ion, z[n] was low-pass il e ed a a cu -o equency o c2= 0.6 Hz aiming o66
supp ess luc ua ions due o CCs as well as high equency noise. In his case, he signal ob ained67
is exp essed as z [n].68
Peak de ec ion and ea u e ex ac ion69
The zc[n] and z [n] signals we e analyzed independen ly o de ec he ins an s o he local70
maxima. Fo each local maximum de ec ed in zc[n] and z [n], which migh be a po en ial CC o 71
en ila ion espec i ely, he ollowing wa e o m ea u es we e compu ed:72
•A1:The ough- o-peak ampli ude o he luc ua ion.73
•A2:The peak- o- ough ampli ude o he luc ua ion.74
•d1:The du a ion o he ough- o-peak ise.75
•d2:The du a ion o he peak- o- ough all.76
The ea u es and loca ions o each luc ua ion in zc[n] and z [n] we e s o ed in he ec o s77
cand espec i ely (Fig. 2). Panel b in Fig. 1 illus a es, om op o bo om espec i ely,78
examples o z[n], zc[n] and z [n] signals, whe e he ex ac ed ea u es a e depic ed.79
Decision algo i hm80
Fo CC de ec ion, he decision algo i hm decided whe he a local maximum was conside ed as81
CC, based on he ex ac ed ea u es and loca ion o each luc ua ion, c. The algo i hm e alua ed82
he du a ion o he luc ua ion (d1+d2) agains a s a ic h eshold and he mean ampli ude, mean83
alue o A1and A2, agains a dynamic h eshold gi en by:84
Thc=Wc·
Nc
X
k=1
A1k+A2k
2(1)
Thc ep esen s he weigh ed a e age o he mean ampli ude o he las 6 de ec ed CCs, and85
Wcdeno es he weigh ing ac o . The luc ua ion was classi ied as CC i i s du a ion and mean86
ampli ude we e abo e he h esholds es ablished by he decision algo i hm. A e ac o y pe iod87
be ween consecu i e CCs was conside ed o a oid possible alse posi i es.88
Fo en ila ion de ec ion, he decision algo i hm classi ied a local maximum as en ila ion based89
on he ex ac ed ea u es and loca ion o each luc ua ion, . The decision algo i hm e alua ed he90
in la ion ime (d1) agains a s a ic h eshold, and he in la ion ampli ude, A1, agains a dynamic91
h eshold de ined as ollows:92
Th =W ·
N
X
k=1
min(A1k, A2k) (2)
Th desc ibes he weigh ed a e age o he minimum ampli ude o he las 17 de ec ed93
en ila ions, and W deno es he weigh ing ac o . The luc ua ion was classi ied as en ila ion94
i i s in la ion ime and in la ion ampli ude we e abo e he h esholds es ablished by he decision95
algo i hm. A e ac o y pe iod be ween consecu i e en ila ions was conside ed in o de o a oid96
possible alse posi i es.97
The ins an s o CCs and en ila ions de ec ed by he decision algo i hm we e deno ed as cand98
espec i ely (Fig. 2).99
2.3. CPR quali y me ics100
Se en di e en CPR quali y me ics we e compu ed using cand ins an s:101
•Mean CC- a e: Mean equency o CCs du ing CC-se ies. Consecu i e CCs sepa a ed by102
less han 1.5 s o med a CC-se ies.19 A single alue pe episode was epo ed.103
•F ac ion o minu es wi h inadequa e CC- a e (FMIR): P opo ion o minu es wi h104
CC- a e below 90 cpm o abo e 120 cpm.19 A ole ance o 10 cpm om he lowe limi 105
(100 cpm) ecommended by he 2010 esusci a ion guidelines1was es ablished o conside 106
a CC- a e as inadequa e. A single alue pe episode was epo ed.107
•Ches comp ession ac ion (CCF): P opo ion o ime wi hou spon aneous ci cula ion108
du ing which CCs we e p o ided.31 A single alue pe episode was epo ed.109
•Mean en ila ion a e: Mean alue o he en ila ions p o ided e e y minu e o he110
episode.19 A single alue pe episode was epo ed.111
Acknowledgemen s257
This wo k ecei ed inancial suppo om he Minis e io de Econom´ıa y Compe i i idad o 258
Spain h ough he p ojec s TEC2012-31928 and TEC2012-31144, om he Uni e si y o he Basque259
Coun y (UPV/EHU) h ough he uni UFI11/16 and om he P og ama de Fo maci´on de Pe sonal260
In es igado del Depa amen o de Educaci´on, Uni e sidades e In es igaci´on del Gobie no Vasco261
h ough he g an s BFI-2010-174, BFI-2010-235 and BFI-2011-166.262
Re e ences263
[1] Nolan JP, Soa J, Zideman DA, e al. Eu opean Resusci a ion Council Guidelines o Resusci a ion 2010 Sec ion264
1. Execu i e summa y. Resusci a ion 2010;81(10):1219–1276.265
[2] Meaney PA, Bob ow BJ, Mancini ME, e al. Ca diopulmona y Resusci a ion Quali y: Imp o ing Ca diac266
Resusci a ion Ou comes Bo h Inside and Ou side he Hospi al A Consensus S a emen F om he Ame ican267
Hea Associa ion. Ci cula ion 2013;128(4):417–435.268
[3] Babbs CF, Voo hees WD, Fi zge ald KR, Holmes HR, Geddes LA. Rela ionship o blood p essu e and low269
du ing CPR o ches comp ession ampli ude: E idence o an e ec i e comp ession h eshold. Annals o 270
Eme gency Medicine 1983;12(9):527–532.271
[4] Fi zge ald KR, Babbs CF, F isso a HA, Da is RW, Sil e DI. Ca diac ou pu du ing ca diopulmona y272
esusci a ion a a ious comp ession a es and du a ions. Ame ican Jou nal o Physiology - Hea and273
Ci cula o y Physiology 1981;241(3):H442–H448.274
[5] Sunde K, Wik L, Naess PA, G und F, Nicolaysen G, S een PA. Imp o ed haemodynamics wi h inc eased275
comp ession-decomp ession a es du ing ACD-CPR in pigs. Resusci a ion 1998;39(3):197–205.276
[6] Chei e z IM, C aig DM, Quick G, e al. Inc easing idal olumes and pulmona y o e dis en ion ad e sely a ec 277
pulmona y ascula mechanics and ca diac ou pu in a pedia ic swine model. C i ical ca e medicine 1998;278
26(4):710–716.279
[7] Milande MM, Hiscok PS, Sande s AB, Ke n KB, Be g RA, Ewy GA. Ches comp ession and en ila ion a es280
du ing ca diopulmona y esusci a ion: he e ec s o audible one guidance. Academic eme gency medicine:281
o icial jou nal o he Socie y o Academic Eme gency Medicine 1995;2(8):708–713.282
[8] Yu T, Weil MH, Tang W, e al. Ad e se Ou comes o In e up ed P eco dial Comp ession Du ing Au oma ed283
De ib illa ion. Ci cula ion 2002;106(3):368–372.284
[9] Au de heide TP, Sigu dsson G, Pi allo RG, e al. Hype en ila ion-induced hypo ension du ing285
ca diopulmona y esusci a ion. Ci cula ion 2004;109(16):1960–1965.286
[10] Wik L, S een PA, Bi che NG. Quali y o bys ande ca diopulmona y esusci a ion in luences ou come a e 287
p ehospi al ca diac a es . Resusci a ion 1994;28(3):195–203.288
[11] Gallaghe EJ, Lomba di G, Gennis P. E ec i eness o bys ande ca diopulmona y esusci a ion and su i al289
ollowing ou -o -hospi al ca diac a es . JAMA: he jou nal o he Ame ican Medical Associa ion 1995;290
274(24):1922–1925.291
[12] Abella BS, Sandbo N, Vassila os P, e al. Ches comp ession a es du ing ca diopulmona y esusci a ion a e292
subop imal: a p ospec i e s udy du ing in-hospi al ca diac a es . Ci cula ion 2005;111(4):428–434.293
[13] Edelson DP, Abella BS, K ame -Johansen J, e al. E ec s o comp ession dep h and p e-shock pauses p edic 294
de ib illa ion ailu e du ing ca diac a es . Resusci a ion 2006;71(2):137–145.295
[14] Wik L, K ame -Johansen J, Myklebus H, e al. Quali y o ca diopulmona y esusci a ion du ing ou -o -hospi al296
ca diac a es . JAMA: he jou nal o he Ame ican Medical Associa ion 2005;293(3):299–304.297
[15] Abella BS, Al a ado JP, Myklebus H, e al. Quali y o ca diopulmona y esusci a ion du ing in-hospi al ca diac298
a es . JAMA: he jou nal o he Ame ican Medical Associa ion 2005;293(3):305–310.299
[16] K ame -Johansen J, Myklebus H, Wik L, e al. Quali y o ou -o -hospi al ca diopulmona y esusci a ion wi h300
eal ime au oma ed eedback: a p ospec i e in e en ional s udy. Resusci a ion 2006;71(3):283–292.301
[17] Pe kins GD, Boyle W, B idges ock H, e al. Quali y o CPR du ing ad anced esusci a ion aining.302
Resusci a ion 2008;77(1):69–74.303
[18] Kampmeie TG, Lukas RP, S e le C, e al. Ches comp ession dep h a e change in CPR guidelines-Imp o ed304
bu no su icien . Resusci a ion 2014;85(4):503–508.305
[19] K ame -Johansen J, Edelson DP, Lose H, K¨ohle K, Abella BS. Uni o m epo ing o measu ed quali y o 306
ca diopulmona y esusci a ion (CPR). Resusci a ion 2007;74(3):406–417.307
[20] Hos le D, E e son-S ewa S, Rea TD, e al. E ec o eal- ime eedback du ing ca diopulmona y esusci a ion308
ou side hospi al: p ospec i e, clus e - andomised ial. BMJ 2011;342.309
[21] Becke s SK, Sko ning MH, F ies M, e al. CPREzy imp o es pe o mance o ex e nal ches comp essions in310
simula ed ca diac a es . Resusci a ion 2007;72(1):100–107.311
[22] Yeung J, Meeks R, Edelson D, Gao F, Soa J, Pe kins GD. The use o CPR eedback/p omp de ices du ing312
aining and CPR pe o mance: A sys ema ic e iew. Resusci a ion 2009;80(7):743–751.313
[23] Sko ning M, Becke s SK, B okmann JC, e al. New isual eedback de ice imp o es pe o mance o ches 314
comp essions by p o essionals in simula ed ca diac a es . Resusci a ion 2010;81(1):53–58.315
[24] Ke n KB, S ickney RE, Gallison L, Smi h RE. Me onome imp o es comp ession and en ila ion a es du ing316
CPR on a manikin in a andomized ial. Resusci a ion 2010;81(2):206–210.317
[25] S eche FS, Olsen JA, S ickney RE, Wik L. T ans ho acic impedance used o e alua e pe o mance o 318
ca diopulmona y esusci a ion du ing ou o hospi al ca diac a es . Resusci a ion 2008;79(3):432–437.319
[26] A amendi E, Ayala U, I us a U, Alonso E, E es øl T, K ame -Johansen J. Supp ession o he ca diopulmona y320
esusci a ion a e ac s using he ins an aneous ches comp ession a e ex ac ed om he ho acic impedance.321
Resusci a ion 2012;83(6):692–698.322
[27] Risdal M, Aase SO, S a land M, E es øl T. Impedance-based en ila ion de ec ion du ing ca diopulmona y323
esusci a ion. IEEE ansac ions on bio-medical enginee ing 2007;54(12):2237–2245.324
[28] Edelson DP, Eile s jønn J, Weidman EK, Re ze E, Hoek TLV, Abella BS. Capnog aphy and ches -wall325
impedance algo i hms o en ila ion de ec ion du ing ca diopulmona y esusci a ion. Resusci a ion 2010;326
81(3):317–322.327
[29] Lose H, Risdal M, S e z F, e al. Tho acic impedance changes measu ed ia de ib illa o pads can moni o 328
en ila ion in c i ically ill pa ien s and du ing ca diopulmona y esusci a ion. C i ical ca e medicine 2006;329
34(9):2399–2405.330
[30] Robe s K, S ini asan V, Niles DE, e al. Does change in ho acic impedance measu ed ia de ib illa o elec ode331
pads accu a ely de ec en ila ion b ea hs in child en? Resusci a ion 2010;81(11):1544–1549.332
[31] Iyanaga M, G ay R, S ephens SW, e al. Compa ison o me hods o he de e mina ion o ca diopulmona y333
esusci a ion ches comp ession ac ion. Resusci a ion 2012;83(5):568–571.334
[32] Id is AH, Daya M, Owens P, e al. High incidence o ches comp ession oscilla ions associa ed wi h capnog aphy335
du ing ou -o -hospi al ca diopulmona y esusci a ion. Ci cula ion 2010;122:A83.336
[33] Ayala U, E es øl T, Alonso E, e al. Au oma ic de ec ion o ches comp essions o he assessmen o 337
CPR-quali y pa ame e s. Resusci a ion 2014, DOI: 10.1016/j. esusci a ion.2014.04.007.338
Figu e Legends339
Figu e 1 Examples o he signals comp ised in each episode: CD, capnog am340
and TI signals om op o bo om in panel a. Re iewe s’ anno a ions341
o CCs and en ila ions a e depic ed as ed do ed lines in CD and342
capnog am signals espec i ely. Panel b shows ( om op o bo om)343
in mo e de ail he segmen colo ed in blue in he TI signal o panel344
a, z[n], he same segmen p ocessed o enhance CCs, zc[n], and he345
same segmen p ocessed o enhance en ila ions, z [n]. The ex ac ed346
ea u es a e also ep esen ed in bo h zc[n] and z [n].347
Figu e 2 Scheme used o compu e he ins an s o bo h CCs and en ila ions348
using exclusi ely he TI signal, z[n].349
Figu e 3 Box plo s showing he pe o mance o he simul aneous350
CC/ en ila ion de ec o . The sensi i i y and PPV alues o 351
CCs and en ila ions o he es se a e depic ed in panels a and b352
espec i ely.353
Figu e 4 Bland-Al man plo s o CC- a e, FMIR, CCF, en ila ion a e and354
FMH a e ep esen ed in a, b, c, d and e panels espec i ely. The 95%355
LOA is depic ed in black dashed lines.356
Figu e 5 Resul s ob ained om he ins an aneous a e analysis. a and357
d panels show he Bland-Al man plo s o ins an aneous CC- a e358
and ins an aneous en ila ion a e espec i ely. The pe cen age359
o ins an aneous a es wi h an e o abo e 10% in CC- a e and360
pe cen age o ins an aneous a es wi h an e o g ea e hen 2 min−1
361
in en ila ion a e a e ep esen ed pe episode in panels b and e. The362
pe cen age o alse a es o ins an aneous CC- a e and ins an aneous363
en ila ion a e a e shown in panels c and espec i ely.364
Figu e 6 (Supp.) Examples o unin e p e able in e als in he capnog am.365
F om op o bo om in e als wi h: ques ionable in la ion/de la ion366
pa e n, CPR a i ac , unusual wa e o m, and CPR a i ac in he367
10 −40 s in e al.368
Figu e 7 (Supp.) Two di e en ime in e als co esponding o he same369
episode a e shown in panels a and b. The CD, z[n], and zc[n] signals370
can be obse ed in bo h panels om op o bo om espec i ely.371
Red do ed lines depic he ins an s o he CCs ma ked in he372
gold s anda d. Red iangles ep esen he ins an s o he co ec ly373
de ec ed CCs ( ue posi i es), whe eas black iangles ep esen he374
ins an s o he inco ec ly de ec ed CCs ( alse posi i es). In zc[n] o 375
panel a, he e o s in he de ec ion due o he second ha monic o he376
TI luc ua ion a e illus a ed. Howe e , he e a e no de ec ion e o s377
in panel b as he e is no second ha monic in he TI.378
16
Figu e 8 (Supp.) Two di e en ime in e als co esponding o he same379
episode a e shown in panels a and b. The capnog aphy, z[n], and z [n]380
signals can be obse ed in bo h panels om op o bo om espec i ely.381
Red do ed lines depic he ins an s o he en ila ions ma ked in382
he gold s anda d. Red iangles ep esen he ins an s o he383
co ec ly de ec ed en ila ions ( ue posi i es), whe eas black iangles384
ep esen he ins an s o he inco ec ly de ec ed en ila ions ( alse385
posi i es). In z [n] o panel a, he e a e no de ec ion e o s.Howe e ,386
he e a e wo de ec ion e o s in panel b due o luc ua ions in he TI387
no co esponding o en ila ions as can be seen in he capnog am.388
CD (mm)
−30
−15
0
Capnog am (mmHg)
0
10
20
Time (s)
a)
TI (Ω)
010 20
104
107
110
z[n] (Ω)
104
107
110
A2
A1
d1
d2
zc[n] (Ω)
103
106
109
Time (s)
b)
A2
A1
d1
d2
z [n] (Ω)
17.11 19.11 21.11
105
106
107
Figu e 1: Examples o he signals comp ised in each episode: CD, capnog am and TI signals om op o bo om
in panel a. Re iewe s’ anno a ions o CCs and en ila ions a e depic ed as ed do ed lines in CD and capnog am
signals espec i ely. Panel b shows ( om op o bo om) in mo e de ail he segmen colo ed in blue in he TI signal
o panel a, z[n], he same segmen p ocessed o enhance CCs, zc[n], and he same segmen p ocessed o enhance
en ila ions, z [n]. The ex ac ed ea u es a e also ep esen ed in bo h zc[n] and z [n].
z[n]
Low-Pass
Fil e
c1
Low-Pass
Fil e
c2
zc[n]
z [n]
Peak
De ec ion
&
Fea u e
Ex ac ion
c
Decision
Algo i hm
Decision
Algo i hm
c
Figu e 2: Scheme used o compu e he ins an s o bo h CCs and en ila ions using exclusi ely he TI signal, z[n].
a)
SENS (%) PPV (%)
60
80
100
b)
SENS (%) PPV (%)
60
80
100
Figu e 3: Box plo s showing he pe o mance o he simul aneous CC/ en ila ion de ec o . The sensi i i y and PPV
alues o CCs and en ila ions o he es se a e depic ed in panels a and b espec i ely.
Gold s anda d (cpm)
Mean CC a ea)
E o (cpm)
90 100 110 120 130
−20
−10
0
10
Gold s anda d ( ac ion)
FMIRb)
E o ( ac ion)
00.2 0.40.60.8
−0.1
0
0.1
0.2
Gold s anda d ( ac ion)
CCFc)
E o ( ac ion)
0.4 0.60.81
−0.05
0
0.05
0.1
Gold s anda d (min−1)
Mean en ila ion a ed)
E o (min−1)
4 8 12 16
−4
−2
0
2
Gold s anda d ( ac ion)
FMHe)
E o ( ac ion)
00.2 0.40.6
−0.04
0
0.04
0.08
Figu e 4: Bland-Al man plo s o CC- a e, FMIR, CCF, en ila ion a e and FMH a e ep esen ed in a, b, c, d and
e panels espec i ely. The 95% LOA is depic ed in black dashed lines.
