Previously, electrocardiography (ECG) screening before anesthesia was often recommended for patients, regardless of their medical histories.1 Currently, the recommendations have changed, but there are varying individual hospital standards such that patient-related (risk) factors, age (>45 to 50 years old), ASA status, and known cardiovascular diseases often necessitate a routine ECG screening before anesthesia.2–4 Because of the increasing economic interests and cost pressures in public health, these recommendations have been reconsidered.5
Interestingly, a large number of ECG pathologies can be detected using only one lead, as PQ, QRS, or QT time periods can be determined in this way. In this manner, atrioventricular (AV) blocks, bundle branch blocks (BB blocks), atrial fibrillation, or long QT syndrome may be diagnosed; these are all relevant unexpected findings in postoperative routine ECG screenings (6.1% BB block, 4.2% AV block, 2.6% atrial fibrillation).6 To reduce costs and expand the possible fields of application, the development of new, innovative technologies should be meaningful and sustainable.
In a previous study, the practicability of an innovative ECG system, consisting of 2 capacitively coupled electrodes (cECG), called the “Aachen ECG chair,” was assessed in volunteers.7 The aim of the present study was to investigate whether a cECG can replace a conventional ECG in some clinical applications. Therefore, we evaluated the reliability and accuracy of a contact-less, capacitive ECG measurement, as opposed to a conventional ECG, using a single lead cECG system. The primary outcomes are the accurate diagnosis of heart rhythms, the accurate measurement of heart rate (HR), and the precise measurements of the PQ, QRS, and QT intervals.
The study was conducted at the University Hospital Aachen after approval by the local ethics committee. Patients from the following 3 departments were recruited after they were given all of the relevant information and provided written informed consent: (1) the anesthesia premedication department, where patients are examined by anesthesiologists before planned surgery; (2) the cardiology outpatient ward, where outpatients with cardiologic diseases, such as intermittent atrial fibrillation or branch blocks, are seen; and (3) the cardiology day ward, where patients are hospitalized before planned electrical cardioversions.
In the premedication department, volunteer patients >50 years old, corresponding to the hospital's standards for applying a routine ECG, were recruited before planned surgery. The outpatient ward was chosen to use the cECG to evaluate a population with a higher incidence of pathological ECGs. To study individual patient-related influence on the cECG, we examined patients recruited in the day ward before and after electrical cardioversion. The exclusion criteria for all clinical settings included pregnancy, the presence of a pacemaker, and implanted cardioverter defibrillators.
ECG Measurement and Signal Processing
All subjects simultaneously underwent cECG and ECG measurements while sitting in a chair (Fig. 1). Capacitively coupled electrodes were integrated into a cushion of the seat, with active shielding for the cECG recording. The electrode location was comparable to the Einthoven lead II, conducted from the back. In addition to the patient's clothing, layers of a highly resistant varnish on the electrodes ensured electrical isolation and, thus, capacitive coupling. Assuming a defined and constant thickness of varnish, the coupling capacity was mainly influenced by the clothing between the electrode and the patient's skin. A common (4-channel) ECG—consisting of leads I, II, III, aVF, aVL, and aVR, according to Einthoven—was taken as a reference, but only lead II was used for comparative purposes.
The amplified and filtered cECG signals (Fig. 1) and the conventional ECG data stream were fed into a common anesthesia monitoring system (Philips IntelliVue MP 70, Philips Medical Systems, Eindhoven, Netherlands) via an additional open interface module. The data were transferred synchronously in real time to a notebook computer and recorded via a custom Labview software application (Labview v. 8.6, National Instruments Inc., Austin, TX). ECG data were exported as a comma-separated values (CSV) file, and further signal processing was performed using a custom Matlab program (MATLAB 2009a, The MathWorks Inc., Natick, MA).
Patient Information and ECG Allocation
The thickness of each clothing layer was measured with calipers, and additional information about textile structures was recorded. Before the ECG recordings, the patients were allocated into 1 of 4 groups (sinus rhythm, atrial fibrillation, myocardial infarction, and branch block) according to their premedical histories. A cardiac history–related collection of personal medical data and presumed causes for impaired signal quality was recorded in an experiment report, when indicated.
ECG Analysis and Statistics
The recorded data were randomized concerning their capacitive or conductive origin. Signal sections of comparable quality were manually extracted by a physician. For these sections, streams of 5 seconds in length for ECG and cECG were separately printed in a common ECG layout (50 mm/s feed) for further analysis.
The diagnosis of the rhythm, the incidence of extrasystoles, and the measurement of HR and PQ, QRS, and QT time intervals from the cECG and ECG recordings were performed by 2 clinicians (an internist and an anesthesiologist). The mean values of the time periods determined by the clinicians were calculated. We excluded time periods, which were not measureable by at least 1 of the clinicians, from the subsequent analysis. Differences were calculated and analyzed using the Wilcoxon test.
The following intervals were used to define the clinical findings: bradycardia (HR <60 beats per minute [bpm]), tachycardia (HR >100 bpm), AV blocks grade I (PQ >200 ms), incomplete BB blocks (QRS duration between 100 and 120 ms), complete BB blocks (QRS duration >120 ms), and long QT syndrome (HR-corrected QT duration >440 ms).
For each measured time period, the difference between cECG and ECG was calculated. Thus, ΔHR, ΔPQ, ΔQRS, and ΔQT were obtained and defined as inaccuracy. As potential impacts on inaccuracy, we examined body weight, body height, body mass index, clothing thickness, and type of textile.
SPSS 17 for Windows (SPSS Inc., IBM Business Analytics Software, Armonk, NY) was used for the statistical analysis of all results, including tests related to normal distribution (Kolmogorov Smirnov test), paired t tests and Spearman correlation. MedCalc (MedCalc Software BVBA, Mariakerke, Belgium) was used for the Bland–Altman plots.
In total, 145 patients were included in this study from the anesthesia premedication department (n = 65), cardiology outpatient ward (n = 66), and cardiology day ward (n = 14). Reliable sequences of sufficient quality and at least 5 seconds in length were extracted from 107 patients. One recording from each patient was used for further analysis (Table 1).
In the premedication department, 23 recordings and in the cardiology outpatient ward, 10 were excluded. The root causes were an unbalanced contact pressure due to a nonupright sitting position, moving artifacts, and partial pathologies of the spine or thorax (e.g., scoliosis, kyphosis, and emphysema). Another problem was noticed in the cardiology day ward, where 5 recordings were excluded; after cardioversion, partially no signal (“zero line”) was derivable.
To examine whether the quantification of the time periods and thus the cECG or ECG diagnosis was significantly affected by subjective factors, the evaluations of the 2 clinicians were compared. The mean differences in the time periods obtained from conventional ECG were inside an assumed measuring error of approximately 30 ms. With cECG, the QRS duration differed significantly (39.8 ms, P < 0.001).
In summary, HRs were evaluable in all cases, PQ times in 85%, QRS duration in 99%, and QT duration in 97% of all reliable cECG recordings. The values for QRS duration differed significantly between cECG and ECG (Wilcoxon signed ranks test, P < 0.001). Inaccuracies for HR, PQ period, QRS duration, and QT period were calculated and correlated with confounding factors (Table 1).
Furthermore, Bland–Altman plots comparing cECG and ECG readings showed the following bias ± limits of agreement: HR (0.4 ± 4.3 per minute), QRS duration (32.7 ± 74.4 ms), PQ (−1.9 ± 50.5 ms), and QT period (4.4 ± 112.6 ms) (Fig. 2).
Generally, high concordance was achieved between cECG and ECG for the diagnostic findings examined; however, the detection of extra systoles and the diagnosis of a long QT syndrome were weak (Table 2). Receiver operating characteristic curves were generated for BB blocks and long QT syndromes (Fig. 3).
Several factors were found to influence the cECG signal. Inaccuracy was lowest in obese patients (Table 1). Nevertheless, according to the Spearman ρ test, this correlation was only a trend (P = 0.05). In cases of nonanalyzable data, the following notes were taken: kyphosis (4 cases), pulmonary emphysema (2 cases), prior spine surgery (2 cases), and extensive sweating (2 cases). The thickness of clothing varied from 0 mm to 3.3 mm, and the mean thickness did not differ between the feasible and excluded recordings (0.64 vs 0.50 mm, P = 0.26); however, the type of textiles did differ. The exclusion rate for patients with single textile materials (cotton or polyester) was 21%; for mixed textiles the exclusion rate was 50%. Signals from patients with naked backs were largely excluded (60%) (Table 1).
In this clinical trial, we evaluated the reliability and accuracy of capacitive ECG electrodes, which were integrated into a seat cushion. ECG measurements were easy to obtain in 74% of all patients. Clinicians used these ECG recordings to accurately measure HR and correctly diagnose sinus rhythm, atrial fibrillation, and AV blocks. High sensitivity was achieved for the diagnosis of tachycardia, bradycardia, atrial fibrillation, and complete BB block. The anesthesia premedication department and the cardiology outpatient ward were identified as practical settings in which to use a cECG device for screening purposes.
Correct electrode placement was performed in each case simply by having the patients sit on the chair; hence, assistance by a qualified employee is not needed. With the integration of electrodes into chairs, seats, and beds, cECG will explore new application fields in which conventional ECG is not applicable. Apart from chairs in waiting rooms, integration into examination tables (e.g., in the operating room or ambulances), car seats, and nursing beds are potential further applications.
Moreover, compared with conventional single-use electrodes, cECG electrodes are reusable. Considering the relatively low cost of <20 EUR for the electrode components, a favorable cost–benefit ratio can be assumed.
Several limitations and problems of this study must be addressed. Approximately 26% of the sequences did not contain evaluable data. In some of these cases, it was impossible to obtain a feasible signal because of movement artifacts or patient pathologies. Kyphosis and thorax deformations caused by pulmonary emphysema, which led to inadequate coverage of the electrodes, were the most substantial obstacles to obtain a cECG of sufficient signal quality.
A deformation of the signal waveform (Fig. 4) affected the QRS waveform in several cECG recordings. Therefore, diagnoses of BB blocks and long QT syndromes were hampered. For long QT syndromes, the sensitivity and specificity were moderate (Table 2), whereas incomplete BB blocks were detected with moderate sensitivity but poor specificity. The signal deformations were presumably caused by the high-pass-filtering nature of the electrodes. In theory, a correcting factor that compensates for these overshoots could be developed.
Bodyweight was identified as another factor influencing the cECG signal quality. The mean body weight of patients with reliable cECG signals was significantly higher (82.6 vs 74.8 kg, P = 0.02), possibly due to the increased surface pressure and an expanded area of body contact with the electrodes. A first approach for a tailored positioning of the electrodes in the majority of patients was already described by Eilebrecht et al. in a follow-up project.8
Furthermore, several studies have described the susceptibility of cECG devices to motion artifacts below 0.8 Hz and above 10 Hz.9,10 As can be shown analytically, signal deformations, particularly of the QRS complex, might be caused by variations in electrode coupling.aFigure 2 suggests that an overestimation of QRS duration is related to the length of the QRS complex. One explanation for this finding could be that clear identification of the zero-crossing point with the equipotential line of the QRS complex, both at the beginning and the end, was difficult in some cases.
An impedance measurement could be beneficial for the characterization of the coupling between the electrodes and the patient. This measurement could be used as a quality marker for coupling or for the compensation for movement artifacts.
These expansions and modifications will not solve all the problems encountered in this study, but will certainly improve the reliability of cECG recordings, especially regarding movement artifacts.
Several interesting findings were noted concerning the impact of clothing materials. However, a very detailed analysis of clothing was not possible due to the small sample size. We intrinsically expected an excellent signal quality for cECG recordings in patients with naked backs, but the exclusion rate was highest for this subpopulation. Remarkably, after cardioversion intensive sweating degraded the capacitive coupling. Apparently, the low conductance hindered the accumulation of electrical charge around the electrodes. A short circuit led to an immediate discharge, thus causing a zero line cECG.
Another factor for a high exclusion rate was mixed-fiber textiles, possibly due to triboelectric effects.11 Additionally, electrostatic charges in the surroundings of the patients have been identified as having a considerable impact.
The identification of relevant factors influencing cECG measurements has already been the subject of laboratory simulation trials. In particular, the impact of wet shirts, leading to short circuits under extensively wet conditions, could be reconfirmed (unpublished data of the authors). The relative variation in the coupling capacitances between the patient and the electrodes caused by different factors (e.g., diverse textiles) was identified as the central issue.a,11
Although waveform deformation led to a hampered cECG diagnosis in several recordings, 2 clinicians from different medical departments generally obtained identical results for the derived time periods.
In summary, cECG recording is a promising tool for screening purposes, but it is not yet ready to replace conventional ECG. However, because of the quick and easy use and lower cost of cECG, clinical waiting rooms may be considered fields of application as the detection of heart rhythm and rate in this setting is certainly feasible.
Name: Michael Czaplik, MD.
Contribution: This author helped design the study, conduct the study, acquire the data, analyze the data, and prepare the manuscript.
Name: Benjamin Eilebrecht.
Contribution: This author helped acquire the data, analyze the data, prepare the manuscript, and provided technical support.
Name: Rafael Walocha, MD.
Contribution: This author helped recruit the patients, acquire the data, analyze the data, and provided educational support.
Name: Marian Walter, PhD.
Contribution: This author helped prepare the manuscript and provided technical support.
Name: Patrick Schauerte, MD.
Contribution: This author helped recruit the patients and prepare the manuscript.
Name: Steffen Leonhardt, MD, PhD.
Contribution: This author helped prepare the manuscript and provided technical support.
Name: Rolf Rossaint, MD.
Contribution: This author helped prepare the manuscript and provided technical support.
This manuscript was handled by: Dwayne Westenskow, PhD.
a Eilebrecht B, Czaplik M, Wartzek T, Schauerte P, Leonhardt S. Analysis of influences on capacitive ECG measurements based on a closed loop model. Proceedings 6th ESGCO. Berlin, Germany: 2010.
1. Christian KW, Gervais H, Dick W. Value of preoperative screening studies. Anaesthesist 1988;37:694–703
2. Callaghan LC, Edwards ND, Relly CS. Utilisation of the pre-operative ECG. Anaesthesia 1995;50:488–90
3. Zwissler B; American Hospital Association, American College of Cardiology. Preoperative evaluation of cardiac risk patients for non-cardiac surgery. New guidelines of the ACC/AHA 2007. Anaesthesist 2008; 57: 717–9
4. Clendenin DJ, Athiraman U, Zurakowski D, Shapiro F, Sethna NF. Accuracy of preoperative electrocardiographic and chest radiographic screening for prediction of left ventricular dysfunction in patients with suspected neuromuscular disorders. Anesth Analg 2010;110:1116–20
5. Chung F, Yuan H, Yin L, Vairavanathan S, Wong DT. Elimination of preoperative testing in ambulatory surgery. Anesth Analg 2009;108:467–75
6. Correll DJ, Hepner DL, Chang C, Tsen L, Hevelone ND, Bader AM. Preoperative electrocardiograms: patient factors predictive of abnormalities. Anesthesiology 2009;110:1217–22
7. Czaplik M, Eilebrecht B, Ntouba A, Walter M, Schauerte P, Leonhardt S, Rossaint R. Clinical proof of practicability for an ECG device without any conductive contact. Biomed Tech (Berl) 2010;55:291–300
8. Eilebrecht B, Schommartz A, Wartzek T, Walter M, Leonhardt S. Multichannel ECG-Measurement-System with Capacitive Patient Coupling, 44. Rostock, Germany: Jahrestagung der Deutschen Gesellschaft für Biomedizinische Technik, 2010
9. Such O. Motion tolerance in wearable sensors—the challenge of motion artifact. Conf Proc IEEE Eng Med Biol Soc 2007;1542–5
10. Lee SM, Kim KK, Park KS. Wavelet approach to artifact noise removal from capacitive coupled electrocardiograph. Conf Proc IEEE Eng Med Biol Soc 2008;2944–7
© 2012 International Anesthesia Research Society
11. Wartzek T, Lammersen T, Eilebrecht B, Walter M, Leonhardt S. Triboelectricity in capacitive biopotential measurements. IEEE Trans Biomed Eng 2011;58:1268–77