University Hospital Schleswig-Holstein, Institute for Emergency Medicine, Kiel, GermanyMedical University of Graz, Department of Anesthesiology and Intensive Care Medicine, Division of Anesthesiology for Cardiovascular and Thoracic Surgery and Intensive Care Medicine, Graz, Austria
Search for articles by this authorWolfgang J. Kern 1
Author Footnotes
1 Both authors contributed equally to this paper.
Wolfgang J. Kern
Footnotes
1 Both authors contributed equally to this paper.
Affiliations
University of Graz, Institute of Mathematics and Scientific Computing, Graz, AustriaBioTechMed-Graz, Graz, Austria
Search for articles by this authorBirgitt Alpers
Birgitt Alpers
Affiliations
University Hospital Schleswig-Holstein, Institute for Emergency Medicine, Kiel, Germany
Search for articles by this authorMichael Schörghuber
Michael Schörghuber
Affiliations
Medical University of Graz, Department of Anesthesiology and Intensive Care Medicine, Division of Anesthesiology for Cardiovascular and Thoracic Surgery and Intensive Care Medicine, Graz, Austria
Search for articles by this authorAndreas Bohn
Andreas Bohn
Affiliations
City of Münster Fire Department, Münster, GermanyUniversity Hospital Münster, Department of Anesthesiology, Intensive Care and Pain Medicine, Münster, Germany
Search for articles by this authorMartin Holler
Martin Holler
Affiliations
University of Graz, Institute of Mathematics and Scientific Computing, Graz, AustriaBioTechMed-Graz, Graz, Austria
Search for articles by this authorJan-Thorsten Gräsner
Jan-Thorsten Gräsner
Affiliations
University Hospital Schleswig-Holstein, Institute for Emergency Medicine, Kiel, GermanyUniversity Hospital Schleswig-Holstein, Department of Anaesthesiology and Intensive Care Medicine, Kiel, Germany
Search for articles by this authorJan Wnent
Jan Wnent
Affiliations
University Hospital Schleswig-Holstein, Institute for Emergency Medicine, Kiel, GermanyUniversity Hospital Schleswig-Holstein, Department of Anaesthesiology and Intensive Care Medicine, Kiel, GermanyUniversity of Namibia, School of Medicine, Windhoek, Namibia
Search for articles by this authorShow footnotesHide footnotes
Author Footnotes
1 Both authors contributed equally to this paper.
Open AccessPublished:January 03, 2022DOI://doi.org/10.1016/j.resuscitation.2021.12.028
Chest compression fraction calculation: A new, automated, robust method to identify periods of chest compressions from defibrillator data – Tested in Zoll X Series
Previous ArticleOutcomes of out-of-hospital cardiac arrests after a decade of system-wide initiatives optimising community chain of survival in Taipei city
Next ArticleSerum levels of the cold stress hormones FGF21 and GDF-15 after cardiac arrest in infants and children enrolled in single center therapeutic hypothermia clinical trials
To introduce and evaluate a new, open-source algorithm to detect chest compression periods automatically by the rhythmic, high amplitude signals from an accelerometer, without processing single chest compression events, and to consecutively calculate the chest compression fraction (CCF). A consecutive sample of defibrillator records from the German Resuscitation Registry was obtained and manually annotated in consensus as ground truth. Chest compression periods were determined by different automatic approaches, including the new algorithm. The diagnostic performance of these approaches was assessed. Further, using the different approaches in conjunction with different granularities of manual annotation, several CCF versions were calculated and compared by intraclass correlation coefficient (ICC). 131 defibrillator recordings with a total duration of 5755 minutes were analysed. The new algorithm had a sensitivity of 99.39 (95% CI 99.38, 99.41)% and specificity of 99.17 (95% CI 99.15; 99.18)% to detect chest compressions at any given timepoint. The ICC compared to ground truth was 0.998 for the new algorithm and 0.999 for manual annotation, while the ICC of the proposed algorithm compared to the proprietary software was 0.978. The time required for manual annotation to calculate CCF was reduced by 70.48 (22.55, [94.35, 14.45])%. The proposed algorithm reliably detects chest compressions in defibrillator recordings. It can markedly reduce the workload for manual annotation, which may facilitate uniform reporting of measured quality of cardiopulmonary resuscitation. The algorithm is made freely available and may be used in big data analysis and machine learning approaches. Cardiac arrest is the third leading cause of death in the western world 1. European Resuscitation Council Guidelines 2021: Epidemiology of cardiac arrest in Europe. Resuscitation. 2021; 161: 61-79//doi.org/10.1016/j.resuscitation.2021.02.007Abstract
Aim
Methods
Results
Conclusion
Abbreviations:
CCF (chest compression fraction), CPR (cardiopulmonary resuscitation), ICC (intraclass correlation coefficient), IQR (interquartile range), MCC (Matthews correlation coefficient), SD (standard deviation), 95%-CI (95% confidence interval)Keywords
Introduction
2.
- Gräsner J.-T.
- Wnent J.
- Herlitz J.
- et al.
Survival after out-of-hospital cardiac arrest in Europe -– Results of the EuReCa TWO study.
Resuscitation. 2020; 148: 218-226//doi.org/10.1016/j.resuscitation.2019.12.042
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (228)
- Google Scholar
3.
- Kramer-Johansen J.
- Edelson D.P.
- Losert H.
- Köhler K.
- Abella B.S.
Uniform reporting of measured quality of cardiopulmonary resuscitation (CPR).
Resuscitation. 2007; 74: 406-417//doi.org/10.1016/j.resuscitation.2007.01.024
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (165)
- Google Scholar
The duration of pauses is commonly expressed through chest compression fraction (CCF), where the latter is defined as the quotient of the overall time with ongoing chest compressions and the entire duration of the resuscitation episode, where a pulse is deemed to be absent.
Two parameters are commonly used to measure and quantify chest compressions in defibrillator records. One is thoracic impedance as an indirect measurement of changes in the conductivity of the chest due to chest compressions.
4.
- Stecher F.S.
- Olsen J.-A.
- Stickney R.E.
- Wik L.
Transthoracic impedance used to evaluate performance of cardiopulmonary resuscitation during out of hospital cardiac arrest.
Resuscitation. 2008; 79: 432-437//doi.org/10.1016/j.resuscitation.2008.08.007
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (75)
- Google Scholar
5.
- Abella B.S.
- Alvarado J.P.
- Myklebust H.
- et al.
Quality of cardiopulmonary resuscitation during in-hospital cardiac arrest.
JAMA. 2005; 293//doi.org/10.1001/jama.293.3.305
- Crossref
- PubMed
- Scopus (1020)
- Google Scholar
6.
- Wik L.
- Kramer-Johansen J.
- Myklebust H.
- et al.
Quality of cardiopulmonary resuscitation during out-of-hospital cardiac arrest.
Jama. 2005; 293: 299-304//doi.org/10.1001/jama.293.3.299
- Crossref
- PubMed
- Scopus (1058)
- Google Scholar
7.
- Malanga B.
- Geheb F.J.
Emerging technologies: A primer to real CPR Help® technology.
ZOLL Medical Corporation, 2007
- Google Scholar
Although methods using these parameters are technically well-established, they have not been widely utilised until now. One obstacle hindering widespread usage of defibrillator data might be the difficulty to extract them.
8.
- Nielsen A.M.
- Rasmussen L.S.
Data management in automated external defibrillators: a call for a standardised solution.
Acta Anaesth Scand. 2011; 55: 708-712//doi.org/10.1111/j.1399-6576.2011.02454.x
- Crossref
- PubMed
- Scopus (8)
- Google Scholar
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
Gupta et al. reasoned that, to obtain a valid representation of CCF, it is acceptable to limit manual annotations to marking the occurrence of (re-)arrest and either return of spontaneous circulation (ROSC) or termination of CPR, which corresponds to determining the denominator of the CCF. For the numerator, they suggested to use single chest compressions as identified by the defibrillator software without manual correction.
Nevertheless, this potentially leads to inaccuracy in CCF because of the technical shortcomings to automatically identify each chest compression event reliably. Misidentification of chest compressions occurs due to artefacts in the thoracic impedance signal or accelerometers. Either movement of the patient is falsely identified as chest compression, or chest compressions are not identified, as the signal does not reach a specific threshold. Furthermore, the detected chest compression may also be time-shifted due to artefacts, resulting in a gap, which can be falsely identified as a pause.
Although it is challenging to identify and quantify a single chest compression from thoracic impedance or acceleration signals correctly, the rhythmic, high amplitude signal resulting from ongoing chest compressions allows differentiation of periods with continuous chest compressions clearly from pauses, without deriving compression depth.
We hypothesised that an automatic differentiation of pauses from chest compression periods can be achieved more accurately by identifying periods with sequences of chest compressions directly, rather than identifying single chest compression events first and deriving periods with chest compressions second.
This study aimed to develop a robust, open-sourced algorithm to compute CCF with minimal workload for manual annotation and independent of proprietary software.
Methods
A consecutive sample of cases with defibrillator records for the entire year 2020 was obtained from the German Resuscitation Registry.
A python-based signal processing algorithm was developed to identify chest compression periods in accelerometer signals of these records (Python v.3.7.3, The Python Software Foundation). The algorithm is based on a sliding window approach analysing the root mean square power of the acceleration signal without the intention to detect single chest compression cycles. For ease of understanding, the new algorithm is described in prose form in a data co-submission.
10.
Kern WJ, Orlob S, Alpers B, et al. A sliding-window based algorithm to determine the presence of chest compressions from acceleration data. Data in Brief – Co-submission.
- Google Scholar
In order to evaluate the proposed algorithm, the following steps were carried out:
- (1)
Manual annotation of defibrillator records to obtain a ground truth.
- (2)
Implement different methods to determine chest compression periods and compare them to the ground truth.
- (3)
Implement different methods for computing the resulting CCF and compare them to each other.
These steps are described in more detail in the following.
Step 1 – Manual annotation
Four independent physician-researchers (SO, JW, MS, BA), all trained and experienced in advanced life-support, annotated the cases via an interactive, web-based plotting tool (The Littlest JupyterHub v.0.1, Project Jupyter; iPywidgets v.7.6.3, IPython project).
11.
- Perez F.
- Granger B.E.
Ipython: A system for interactive scientific computing.
Comput Sci Eng. 2007; 9: 21-29//doi.org/10.1109/mcse.2007.53
- Crossref
- Scopus (0)
- Google Scholar
The manual annotation process was subdivided into three subtasks with advancing levels of granularity:
- •
1st task – setting denominator of CCF: The physiological state (cardiac arrest vs supposed spontaneous circulation) and the resuscitation state (resuscitation measures taken vs resuscitation withheld) were annotated as intervals.
This task corresponds to “limited annotation” by Gupta et al.
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
- •
2nd task – setting numerator of CCF: Periods of chest compressions were denoted as intervals, separated by pauses. Specific definitions were given to the annotators. As such a period of chest compressions was defined as at least three successive chest compressions. The absence of chest compressions was identified as a pause when the corresponding episode lasted longer than the duration of two successive chest compressions.
This annotation task was used as ground truth for chest compression periods.
- •
3rd task – revising chest compressions: Every chest compression event identified by the manufacturer software was reviewed and eventually corrected. The manual annotation was irrespective of chest compression qualities like compression depth, rate, or full recoil. Shallow chest compressions not detected by the manufacturer were added manually.
In conjunction with the 1st task, this task corresponds to the classic full annotation, implemented in the workflow of common manufacturers software as CODE-STAT Reviewer (v11.0, Stryker, Kalamazoo, Michigan, United States) and thereby resembles the “manual annotation” by Gupta et al.
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
Two independent researchers annotated each recording separately for the first two tasks. Deviations of more than 0.5 seconds between the markers of the two researchers were automatically flagged and corrected in consensus. The last annotation task was done by a single researcher (SO). The required time for annotation in every task was tracked automatically.
Step 2 – Chest compression period determination
By design, the proposed algorithm and the consensus (2nd task of manual annotation) provide periods of chest compressions.
In addition, two methods were implemented to obtain periods of ongoing chest compressions from the list of single chest compression events detected by the manufacturer software.
The first method checked the temporal difference between chest compression events and decided whether they were continuous (“cycle duration analysis”). The exact definitions given to the annotators were applied.
The second approach was a binning method checking the presence of at least one chest compression for every second of recording, regardless of the continuity (“binning method”). This method is in accordance with the software RecueNet Code Review (v.5.9.0.5, ZOLL Medical Corporation, Chelmsford, Massachusetts, United States).
The performance of different methods to determine chest compressions periods was assessed by binarily classifying a timepoint at every 0.2 seconds of each recording as either chest compressions present or absent, and contrasting it with ground truth (2nd task manual annotation). We then calculated several performance indices (sensitivity, specificity, F1-score
12.
- Dice L.R.
Measures of the amount of ecologic association between species.
Ecology. 1945; 26: 297-302//doi.org/10.2307/1932409
- Crossref
- Google Scholar
13.
- Matthews B.W.
Comparison of the predicted and observed secondary structure of T4 phage lysozyme.
Biochimica et Biophysica Acta BBA – Protein Struct. 1975; 405: 442-451//doi.org/10.1016/0005-2795(75)90109-9
- Crossref
- PubMed
- Scopus (3402)
- Google Scholar
Step 3 – Computation of CCF
In order to ensure comparability with previous publications and existing manufacturer software, six different methods for CCF calculation were implemented. The six different methods correspond to combinations of I different methods to determine periods of ongoing chest compressions (numerator of CCF), II different granularities of manual annotation, and III two different values for the cardiac arrest interval (denominator of CCF). A summary of the different methods for CCF calculation is provided in Table 1. The deviation from CCFground truth was calculated for each method. In addition, intraclass correlation coefficients between all pairs of different CCF methods were computed.
14.
- Liljequist D.
- Elfving B.
- Roaldsen K.S.
Intraclass correlation – A discussion and demonstration of basic features.
Plos One. 2019; 14e0219854//doi.org/10.1371/journal.pone.0219854
- Crossref
- PubMed
- Scopus (169)
- Google Scholar
Table 1Different types of CCF calculation based on different origins of the numerator (=period length with chest compressions), the denominator (=cardiac arrest episode), and methods of period quantification. Note that CCFmanual, CCFlimited, and CCFsoftware annotation correspond to those given by Gupta et al. 9. Software annotation of defibrillator files: Ready for prime time?. Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
- Open table in a new tab
The manual workload to obtain each CCF was calculated as the sum of time required to annotate the respectively used tasks of manual annotation.
Statistics
Statistic computations were performed using SciPy (v1.7.3, The SciPy community).
15.
- Virtanen P.
- Gommers R.
- Oliphant T.E.
- et al.
SciPy 1.0: fundamental algorithms for scientific computing in Python.
Nat Methods. 2020; 17: 261-272//doi.org/10.1038/s41592-019-0686-2
- Crossref
- PubMed
- Scopus (8417)
- Google Scholar
This study was approved by the ethics committee of the University of Kiel (Ref. no.: D 421/21) and the scientific advisory board of the German Resuscitation Registry (Ref. no.: AZ 2021-03).
Results
Defibrillator records
Defibrillator records from 139 consecutive adult resuscitation attempts, all ZOLL X Series (ZOLL Medical Corporation, Chelmsford, Massachusetts, United States), were obtained from the German Resuscitation Registry. Eight cases were excluded from analysis as the recording did not cover CPR episodes, erroneous recording, or corrupted files. The total recording duration was 3 days, 23 hours and 55 minutes. The total observed cardiac arrest duration was 2 days, 22 hours and 17 minutes. The median recorded resuscitation duration per case was 18 (20, [1; 79]) minutes.
Step 1 – Manual annotation
Overall, 185 arrest episodes were identified by the annotators with 3385 periods of chest compressions.
Step 2 – Chest compression period determination
In total, 97.9% of the start and end points of chest compression periods that matched in a greedy search between manual annotation and the proposed automatic detection had a deviation of less than 1 second (6466 of 6602). The mean deviation of start and end points of chest compression periods obtained by manual annotation and by detection with our algorithm was −0.029 ± 0.15 seconds and 0.021 ± 0.133 seconds, respectively, which is below the temporal resolution of our manual annotation.
Overall, the proposed algorithm had a sensitivity of 99.39 (95% CI 99.38, 99.41)% and a specificity of 99.17 (95% CI 99.15; 99.18)% for detecting the presence of chest compressions at any given timepoint in the recording. Concomitantly, the F1-score for the algorithm was 0.992 and the MCC was 0.985. Performance indices for all approaches to determine chest compression periods are given in Table 2. Both methods based on the single chest compression event detection (cycle duration analysis and binning method) were less sensitive and less specific.
Table 2Performance indices for different approaches to identify chest compression periods compared to ground truth from the manual annotation.
Interval determination by new algorithmInterval determination based on single chest compression detectionWith cycle duration analysisWith binning methodSensitivity [% +95%-CI]99.39 [99.38; 99.41]98.88 [98.86; 98.90]98.93 [98.90; 98.95]Specificity [% +95%-CI]99.17 [99.15; 99.18]94.10 [94.05; 94.15]93.65 [93.60; 93.70]F1-Score0.9920.9620.960MCC0.9850.9280.924- Open table in a new tab
Step 3 – Computation of CCF
The six different CCFs for every case are visualised in Fig. 1. The CCF calculated by the new algorithm deviated in median by −0.063 (0.481, [−2.221; 2.154])% from the ground truth. The CCF calculation without any manual annotation (CCFsoftware annotation) deviated in median by −37.842 (44.362, [−86.346; 9.155])% from the ground truth, while CCFmanual, CCFlimited, and CCFZOLL overestimated the CCF in median by 0.156 (0.356 [−1.532; 2.053])%, 0.423 (0.878 [−12.56; 4.557])% and 0.941 (1.121, [−12.051; 5.231])%, respectively. ICCs between the CCF calculation methods are given in Table 3. Compared to CCFground truth, the ICC of the different methods was 0.999 for manual annotation, 0.998 for the new algorithm, 0.984 for limited annotation and 0.978 for the method in analogy to the manufacturer software (CCFZOLL). CCFsoftware annotation, being the method irrespective of the cardiac arrest interval, had an ICC of 0.103 with CCFground truth.
Fig. 1Scatter plots of chest compression fractions for each case, calculated by the different methods. The upper subplot displays absolute values of CCF, while the lower shows the deviation from ground truth for each calculation method. Cases are sorted by the ground truth value. Colour and marker configuration resemble the calculation method; the legend applies to both subplots. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
- View Large Image
- Figure Viewer
- Download Hi-res image
- Download (PPT)
Table 3Crosstable of intraclass correlation coefficients for different methods of chest compression fraction calculation.
CCF methodGround truthNew algorithmManualLimitedSoftware annotationZOLLGround truth10.9980.9990.9840.1030.978New algorithm0.99810.9980.9840.1080.976Manual0.9990.99810.9860.1030.98Limited0.9840.9840.98610.0990.994Software annotation0.1030.1080.1030.09910.091ZOLL0.9780.9760.980.9940.0911- Open table in a new tab
Manual workload
The required time for manual annotation to compute a CCF by the respected method is illustrated in Fig. 2 as a boxplot. With a median duration of 77 (87, [19; 596]) seconds, the workload for annotation for CCFnew algorithm, CCFZOLL, and CCFlimited was the same by design, being the time required to annotate the denominator in task 1. The time required to obtain the ground truth by manually annotating chest compression periods, and the 1st task required a median duration of 309 (331, [39; 1997]) seconds. This duration was not significantly different to the workload required to calculate CCFmanual, i.e., annotation tasks 1 and 3, with a median duration of 300 (380 [49; 1706]) seconds. Thereby automation of chest compression period detection decreased the time required for annotation by 70.48 (22.55, [94.35, 14.45])%.
Fig. 2Boxplot of the required times for manual annotation to calculate the respective chest compression fraction. Asterisks indicate significant differences between the methods. Note that the left (blue) box represents the time for manual annotation to calculate CCFnew algorithm, CCFZOLL, and CCFlimited, which was the same by design, being the time required to annotate the denominator in task 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
- View Large Image
- Figure Viewer
- Download Hi-res image
- Download (PPT)
Discussion
This study addresses two problems: the identification of chest compression periods and the subsequent calculation of CCF.
As a solution, we proposed a standardised algorithm to automatically classify defibrillator recordings' subsegments into those with and those without chest compressions, based on the accelerometer signal. The new algorithm accomplished this with exceptional performance.
The proposed algorithm can be utilised by other research groups and integrated by manufacturers into proprietary software. To our knowledge, such a method has not been described until now. However, relating the presence of chest compression to the occurrence of several chest compressions in a sliding window was described previously in general.
7.
- Malanga B.
- Geheb F.J.
Emerging technologies: A primer to real CPR Help® technology.
ZOLL Medical Corporation, 2007
- Google Scholar
Classification of periods with chest compressions present in contrast to those without is a basic but crucial task for any other processing to analyse defibrillator records. This applies to CCF calculation, as in the present study, but especially to all future machine learning endeavours.
Interestingly, previous workflows to calculate CCF focused on identifying single chest compression events on a microscopic level instead of determining the general presence of ongoing chest compressions on a macroscopic level. Because of the microscopic approach, established methods for example misinterpret the muscle twitch by defibrillation as chest compression, while shallow chest compressions are neglected. Thus, identifying single chest compression events to quantify chest compression periods resembles an unnecessary extra procedure, with additional potential for error, as reflected by our data. Those detection failures have to be corrected by labour-intensive manual annotation. It shall be mentioned that the software by ZOLL does not allow for correction of single chest compression events, while other manufacturer solutions like the one by Stryker (previously PhysioControl) do.
In contrast, our new method identifies chest compressions by the rhythmic, coherent presence of accelerations with high amplitudes in contrast to pauses (Fig. 3). The higher specificity reflects the robustness of our algorithm regarding artefacts compared to the period detection based on single chest compressions. Hence, our algorithm even can be used as a prefilter before analysing other single chest compression qualities like compression depth, recoil, et cetera.
Fig. 3Illustration of different chest compression period detection approaches in two CPR sequences. The purple line represents the accelerometer signal, the derived chest compression events (by the manufacturer software) are represented by purple dots. A grey rectangle highlights each other second in time. The bold lines illustrate the different pause detection approaches (green: ground truth form consensus, blue: new algorithm, orange: binning method as used in proprietary software), a line on the baseline represents the detection of a pause. Left subplot: Note how missed chest compression detection by the manufacturer software causes pauses of one-second length by the binning method (ZOLL) due to the extra task of individual chest compression event detection. Right subplot: Note how a falsely identified chest compression at the beginning of the pause shortens the pause by more than one second due to binning, while the binning also shortens the pause at the end of the pause. Additionally, an artefact within the pause is equally misidentified as a chest compression and shortens the pause by another second. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
- View Large Image
- Figure Viewer
- Download Hi-res image
- Download (PPT)
Another relevant finding of our study is that the method to transfer a list of time points of single chest compression events into a chest compressions period is important. A binning method, like the one implemented into the software RecueNet Code Review by ZOLL, overestimates the CCF.
Regardless of the method to determine chest compression periods, the denominator must be annotated manually to calculate meaningful CCF. Not determining the denominator and subsequently calculating the CCF for the whole recording duration is pointless. This finding is in accordance with the results by Gupta et al.
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
When the manually determined cardiac arrest interval is used as the denominator of CCF, all methods had a good correlation with the ground truth. Nevertheless, we strive to achieve modest marginal gains in high-performance CPR and tools are needed to detect these gains. Therefore, imprecisions of CCF calculation are meaningful.
The CCFs by ground truth, manual annotation and our algorithm correlated exceptionally well. Hence, the annotation process for the numerator can be automated by our algorithm. In accordance with Gupta et al., we found a reduction of time required for manual annotation by 70% due to automatic detection of the numerator of CCF.
9.
- Gupta V.
- Schmicker R.H.
- Owens P.
- Pierce A.E.
- Idris A.H.
Software annotation of defibrillator files: Ready for prime time?.
Resuscitation. 2021; 160: 7-13//doi.org/10.1016/j.resuscitation.2020.12.019
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (4)
- Google Scholar
16.
- Gräsner J.T.
- Herlitz J.
- Koster R.W.
- et al.
Quality management in resuscitation – Towards a European Cardiac Arrest Registry (EuReCa).
Resuscitation. 2011; 82: 989-994//doi.org/10.1016/j.resuscitation.2011.02.047
- Abstract
- Full Text
- Full Text PDF
- PubMed
- Scopus (142)
- Google Scholar
17.
- Jaureguibeitia X.
- Aramendi E.
- Irusta U.
- et al.
Methodology and framework for the analysis of cardiopulmonary resuscitation quality in large and heterogeneous cardiac arrest datasets.