Although clearly valuable, bedside monitoring of cardiac output (CO) and stroke volume (SV) is not used by many intensive care units because of the limitations of currently available techniques. Most of the commonly used continuous CO monitoring devices are invasive.1,2 Due to this invasiveness, the use of pulmonary artery catheters (i.e., the current “gold standard” of CO monitoring) is considered by some to be controversial.3,4 Therefore, less invasive5 or even noninvasive techniques6 have been developed.7 However, the accuracy of continuous CO monitoring based on pulse-contour analysis is reported to be only partially satisfactory.8,9 Also, the noninvasive methods using ultrasound generally require performance by a trained person10 that might be impractical during intensive care when data on CO and SV are frequently required. Therefore, a new noninvasive technique able to monitor CO, particularly when simultaneously providing ventilation monitoring, seems desirable.
Electrical impedance tomography (EIT) is a relatively new method, which shows the regional distribution of changes in electrical impedance within a certain volume with a very high dynamic range (around 50 images/s). Most EIT devices used in various studies only determine changes in electrical impedance regarding a fixed reference; further technical details have been reported by Adler et al.11 and Leonhardt and Lachmann.12 In EIT, mechanical ventilation causes a very strong signal closely correlated with tidal volume, because the electrical impedance of pulmonary tissue is mainly influenced by air content.13 Therefore, this technique is mainly used for ventilation monitoring, for example, to show recruitment of pulmonary tissue.14 Compared with ventilation-related signals in EIT, the signal originating from cardiac activity is >1 order of magnitude smaller. To facilitate cardiac monitoring, the impedance signal should be decomposed into a ventilation-related signal (ZVRS) and a part related to cardiac activity and pulmonary perfusion (ZCPRS) (Fig. 1) to enable cardiovascular monitoring.15 Another possibility is to amplify the perfusion and cardiac-related signal by specific averaging methods (so-called gating) or to temporarily reduce the impedance of blood by IV injection of sodium chloride.
After Barber and Brown16 and Brown et al.17 proposed the use of EIT for clinical applications, McArdle et al.18 and Eyüboglu and Brown19 were among the first to investigate not only regional ventilation but also pulmonary perfusion and cardiac activities. Until now, several methods have been proposed to visualize the relatively small perfusion-related signals using apnea,20,21 cardiac gating,22,23 spectral separation,24–26 or model function-based decomposition such as principle component analysis.27,28
Solà et al.29 used hypertonic saline solution as an indicator, which reduces the impedance of blood to increase the perfusion-related signals, and determined the intrathoracic pulse transit time. Using a similar technique, Borges et al.30 calculated regional pulmonary perfusion (using a 20% NaCl solution as contrast agent) and made a comparison with single-photon emission computed tomography. Apart from these 2 studies, very few publications are available on the quantitative assessment of cardiovascular variables by means of EIT.
We examined the feasibility of EIT for CO monitoring, that is, for robust determination of heart rate (HR) and estimation of changes in SV. In our animal experiments, SV was changed by means of positive end-expiratory pressure (PEEP),31 and a cardiac impedance signal was extracted by means of EIT that was then used to estimate both HR and SV.
Animals and Experimental Procedure
This study was registered and approved by the Ethics Committee of Sahlgrenska University Hospital Gothenburg, Sweden (ID: 179–2010).
For premedication, 15 μg/kg body weight (BW) of dexmedetomidine and 5 mg/kg BW tiletamine and zolazepam were administered IM. Anesthesia was induced with 6 mg/kg BW of pentobarbital sodium and maintained with infusion of pentobarbital sodium 16 mg/kg BW/h and morphine 0.5 mg/kg BW/h.
Due to financial limitations, we chose a sample size of 14 piglets (weighing 37 ± 3 kg) that had their lungs ventilated in pressure-controlled mode in supine position. Due to technical reasons, the datasets of 2 animals were not evaluable. In all experiments, a respiratory rate of 14 breaths/min and an I/E ratio of 1:3 were used. In each animal, 6 single experiments were performed (Fig. 2). In 4 experiments (referred to as PEEP maneuver), PEEP was increased to change SV. Forty-three evaluable PEEP ramp maneuvers were recorded. In the remaining 2 experiments, the HR was increased by infusion of dobutamin to at least 180 bpm at zero PEEP. These latter experiments were mainly used as a plausibility check to evaluate the robustness of the algorithms in HR estimation (schedule in Fig. 2). Before each experiment, the driving pressure was adjusted such that the initial tidal volume equaled 10 mL/kg BW at zero PEEPand was kept constant during each experiment (resulting in a reduced tidal volume during periods of higher PEEP).
Each PEEP experiment consisted of increasing levels of PEEP (ramp maneuver) in which PEEP was set to 0, 5, 10, 15, and 20 cm·H2O (when tolerated). Each setting was maintained for 5 minutes and, during this period, a defined course of actions was performed: after 60 seconds, an end-expiratory apnea of 15 seconds was applied followed by 2 breaths and another inspiratory apnea of 5 seconds. The apnea was performed to collect EIT data without ventilation-related signals and to use these data for validation purposes of the signal-processing algorithm used here (i.e., signal separation into a ventilation-related signal and a cardiac- or perfusion-related signal). Subsequently, SV was measured twice using transpulmonary thermodilution, provided by the calibration procedure of a PiCCO Plus device (Pulsion Medical Systems SE, Feldkirchen, Germany). If both measurements deviated ≥10% from each other, a third measurement was performed and, when indicated, a clearly identifiable erroneous measurement was deleted.
Vital variables monitoring (electrocardiogram, femoral and pulmonary artery pressure, peripheral oxygen saturation, skin temperature) was performed with a Delta Infinity® patient monitoring system (Dräger Medical GmbH, Lübeck, Germany).
To monitor CO and SV (SVTTD), a PiCCO Plus device was used. Although this device provides continuous CO monitoring based on arterial pulse-contour analysis, for the present study we used data from the built-in calibration based on transpulmonary thermodilution measurements. The rationale for this was that, unless it is recalibrated, the device tends to calculate erroneous continuous CO values when there are substantial changes in PEEP. This same device was also used to monitor arterial pressure and SV.
EIT data were recorded using an EIT Evaluation Kit 2 (EEK2; Dräger Medical GmbH, Lübeck, Germany). After shaving the skin, the 16 electrodes embedded in a special silicon waist belt (smaller than belts used for humans) were positioned just below the scapulae, and the reference electrode was attached 10 to 15 cm below the belt on the animal’s right side. Since the proposed analyses were not state-of-the-art, we recorded only the EIT raw data (measured voltages) and processed the data offline using self-developed software implemented in a MatLab® environment (Mathworks, Inc., Natic, MA). Nevertheless, the image reconstruction was calculated by means of proprietary reconstruction algorithms provided by Dräger Medical.
We used an EIT signal decomposition based on principle component analysis, similar to the algorithm proposed by Deibele et al.28 After separating the signals, we obtained a ventilation-related signal (ZVRS) and a cardiac activity and pulmonary perfusion-related signal (ZCPRS). Figure 1 shows 1 cardiac cycle (i.e., 0.75 seconds) of the perfusion-related and cardiac-related signal ZCPRS in 10 steps.
Although there is almost no change in the volume of the heart during the total cardiac cycle, there is a considerable change in the ratio between blood and muscle volume.32 Because of the difference in electrical impedance of blood and cardiac muscle tissue at typical EIT frequencies (50–100 kHz),13 the impedance of the heart is expected to vary during a cardiac cycle. Furthermore, we assume that the change in impedance corresponds with the cardiac SV. This principle was previously shown for a current frequency of 1.3 kHz in an invasive setup.33
To deduce a SV-related signal (ZSV) from the EIT data, we first defined a signal called cardiac impedance ZCard, which is given by the summed blood-related signal (ZCPRS) in a cardiac region, CardROI. The CardROI is outlined in Figure 1, and the mathematical definitions of ZSV and ZCard are given in Equation 1. We assume that the cyclic variation of ZCard correlates with the cardiac SV. Figure 3 shows the change in amplitude of ZCard and global impedance (i.e., the sum of the entire EIT) due to changes in PEEP and indeed supports our hypothesis. Consequently, the cyclic variation of ZCard belonging to the cardiac cycle (cci) is defined as ZSV(cci) at which the index i denotes the number of the cardiac cycle.
Equation (Uncited)Image Tools
Equation 1: Definition of the Cardiac Impedance ZCard and the Derived Parameter ZSV: Note That ZSV Is Calculated for Each Cardiac Cycle (cci) as a Difference Between the Corresponding Diastolic (diai) and Systolic (sysi) Status
Because of the relative nature of the measured impedance signals, all signals denoted with “Z” are given in arbitrary units [AU]. For each experiment (i.e., for the entire PEEP ramp maneuver), the reference is only set once to the least measured impedance. In this way, relative changes of electrical impedance within the chest are consistent during each experiment.
Since this article deals with quantitative assessment of changes in CO and SV, reliable measurements of at least 2 of the following 3 variables are required: HR, changes in SV, and CO. Because EIT provides up to 50 frames/s, estimating HR and SV seems to be most suitable. Therefore, we discuss below the EIT-based estimation of HR and SV.
Since the EIT-based estimation of HR is easy to check, evaluation of HR determination by EIT is also convenient for a plausibility test to determine whether ZCard is in fact cardiac related (as assumed above).
Data are presented as means and standard deviations (SDs). We present bootstrap estimates (medians) and 95% confidence intervals (CIs) (quantile estimates) for between-subject correlations based on the Hamlett multilevel model with additional model building factors cycle (each corresponding to 1 PEEP maneuver) and PEEP settings.34 The bootstrap settings are seed 1567, sample size 10,000, and no stratification. This means that samples are drawn from the whole set of observations. If fitting of the multilevel model failed due to numerical instability, crude correlation coefficients were computed. Accordingly, Pearson correlation coefficients with 95% CIs (via Fisher Z transformation) were calculated for within-subject correlations. Moreover, we calculated weighted Pearson correlation coefficients (by taking means over repeated observations per animal and using these means to calculate the correlation coefficient) with 95% CIs (via Fisher Z transformation, df = 12–1) for between-subject correlations.35 SAS software (V 9.0) for Windows 7 (SAS Institute Inc., Cary, NC) was used for computations.
Besides these statistical evaluations, we propose a further analysis of the SV-related measurements taking into consideration that EIT provides relative data. In this evaluation, a linear regression was calculated between the EIT data ZSV and the reference SVTTD of each maneuver. This yielded a SV estimation called SVEIT in [mL] calculated from ZSV, which can be used to compare the EIT-based estimation of SV and the reference SVTTD. This technique will consider that, in future clinical use, calibration of EIT data is mandatory to estimate any change in volume (tidal breathing volume or SV) based on EIT data. Note that the regression leads to an optimized dataset SVEIT that shows a better correlation with the reference measurement. Therefore, any statistical result based on this analysis is debatable and needs to be considered with caution. Nevertheless, the presented graphs provide a qualitative impression of the feasibility of assessing SV by means of EIT.
Heart Rate Estimation by EIT
Analysis of the HR under PEEP and EIT is based on 81,605 observations in 12 animals. The mean number of observations per maneuver was 6800 (SD 1914.01; range 2022–8726). Animal 1 shows a correlation of 0.466 (95% CI, 0.445–0.485). The correlation coefficients from the remaining animals ranged from 0.92 to 1.00. We calculated the weighted (by sample size) correlation among the animals, yielding a correlation of 0.997 (95% CI, 0.986–0.999) (Fig. 4).
We observed a higher variation between the 2 individual observations for animal 6 in cycle 2 due to an inappropriate positioning of the EIT electrode belt. In animal 4 (cycle 2) and animal 8 (cycle 1), the inaccurate HR estimations are caused by very low arterial blood pressure at the highest PEEP level (the HR reference failed).
When HR was significantly increased by infusion of dobutamin (without any change in PEEP), this resulted in a total of 61,370 observations in the 12 animals [mean 5114.17 (SD 1886.41), range 2556–8994]. The range of correlation coefficients within the repeated observations per animal was between 0.983 to 0.998, resulting in a weighted (by sample size) correlation among the animals of 0.999 (95% CI, 0.996–1.000). All animals showed similar trends over all observations (Fig. 5).a
Variation of Stroke Volume
To analyze the correlation between the ZSV and SVTTD, a multilevel model was fitted to the data. We started with an appropriate model search but found serious instability so that, instead of the sophisticated Roy model,36 only the Hamlett model could be fitted to the data. Five observations had to be excluded from the analysis based on the influence and residual diagnostic. This model was used to establish a bootstrap correlation coefficient (median) and a 95% CI. The correlation coefficient was 0.37 (95% CI, 0.19–0.51). Two animals showed high variation in the data (Fig. 6). Excluding these cases from the analysis results in a (bootstrap median) correlation coefficient of 0.58 (95% CI, 0.43–0.71).
The fact that the EIT data are related to a reference implies that a certain amount of scaling (different in each maneuver) is inherent; this may be one explanation for this moderate correlation. To compensate for the built-in scaling of the EIT data, we calculated a linear regression between ZSV and SVTTD of each maneuver and used the regression to map the ZSV (given in arbitrary units [AU] to [ml]). The mapped SV estimation is defined as SVEIT. Figure 7 presents a scatterplot of the SVEIT and SVTTD based on the same data as presented in Figure 6. Figure 7 shows (qualitatively) the extent to which SVEIT correlates with SVTTD. Although a correlation coefficient of 0.85 (95% CI, 0.78–0.90) can be calculated, these data are statistically unreliable because SVEIT is calculated from SVTTD and ZSV.
However, analyzing the data in more detail reveals a phenomenon that has not been previously reported. The decrease in SV due to an increase in PEEP is expected to cause a decrease ZSV since it is expected to correlate with SV. In contrast, in some of the experiments presented here, we observed an unexpected pattern of the cardiac impedance due to changes in PEEP. Figure 8 shows the continuously calculated ZSV and the global impedance of 4 representative datasets. In all 4 datasets, the reference SVTTD decreased, or at least SVTTD remained unchanged between the 2 adjacent PEEP steps. This phenomenon could be modeled with an intrathoracic environment-related scaling by a factor α of the cardiac impedance ZCard. In some trials, this factor α seems to be a function of lung volume or heart position relative to the electrode plane. We assume that ZSV is given according to Equation 2:
Equation (Uncited)Image Tools
Equation 2: Assumed Multiplicative Model of the Influence on ZSV by the Cardiac Vicinity-Related Factor α
With this equation, we can assign the datasets to 4 categories (C1-C4). In category C1, no change of α during 1 maneuver is visible; in C2, there is a clearly identified amplification of α up to the first increase in PEEP level, and in C3, up to the second increase in PEEP level. In C4, the influence on α seems to be very strong during all PEEP levels. Based on this classification, 32.5% of all the animal experiments belong to C1 and 46.5% belong to C2.
In the present study, we investigated the feasibility of EIT-based cardiac monitoring. To enable cardiac monitoring with EIT, we examined the determination of both HR and SV.
Figure 4 shows that EIT-based estimation of HR using the variations of the cardiac impedance ZCard yields a weighted (by sample size) correlation among the animals of 0.997 (95% CI, 0.986–0.999). During the maneuvers in which HR was significantly changed by dobutamin, the correlation coefficients within the repeated observations per animal ranged from 0.983 to 0.998, resulting in a weighted (by sample size) correlation among the animals of 0.999 (95% CI, 0.996–1.000) (Fig. 5). This high correlation shows that the algorithms provided reliable cardiac-related signals, even with a drastic change in HR or with an exceptionally wide range of ventilation pressures. Interpreting this finding as a plausibility check, we conclude that the signal ZCard is most likely cardiac related (as claimed) and that EIT is able to determine HR in a robust manner.
Estimating the changes in SV by considering the cyclic variations of ZCard (the variations are defined as ZSV) is more challenging. In the present study, we changed SV by applying increasing levels of PEEP from 0 to 20 cm·H2O and monitored SV and EIT. Comparison of ZSV and the reference SVTTD showed only a weak correlation with a correlation coefficient of 0.37 (95% CI, 0.19–0.51). Excluding 2 animals with very high variations resulted in a (bootstrap median) correlation coefficient of 0.58 (95% CI, 0.43–0.71).
Although this correlation might seem relatively low, we have to remember that the EIT data are given in relation to a reference, which causes a linear scaling of the data. Since the reference is different in each maneuver, the weak correlation may be caused by a different scaling of the EIT data. To deal with this, SVEIT was calculated from ZSV and SVTTD using linear regression. In Figure 7, all measurements of SVEIT are presented in comparison to SVTTD; this shows to what extent SVEIT (and ZSV) correlates with SVTTD in each single maneuver if EIT data are calibrated to the current measurement. This evaluation leads to an increased correlation coefficient of 0.85 (95% CI, 0.78–0.90); however, this figure and correlation needs to be considered with caution because SVEIT is adapted to the reference SVTTD. Although this analysis should only be considered qualitatively, it does show that EIT data ZSV might be a valuable parameter to estimate SV continuously if proper calibration is ensured. These results are promising, although even well-established continuous CO monitoring systems show relatively strong deviations from each other (as reviewed by Hadian et al.),37 and errors of 16% are often considered clinically acceptable.38,39
However, establishing a method to calibrate EIT data in SV assessment is not the only challenge for SV monitoring by means of EIT. As shown in Figure 8, an unknown scaling pattern was observed. This scaling α is independent from the inherent scaling mentioned above but seems to be related to lung volume and may depend on the cardiac environment. This dependency might be caused by changes in the heart position relative to the EIT electrode plane or by a changed sensitivity of EIT, as proposed by Patterson et al.40 However, because the present study had no tomographic reference (e.g., computed tomography or magnetic resonance image), we were unable to prove this theory. Although some studies report a robust performance of transpulmonary thermodilution while applying PEEP,41 a second CO reference should be added to the experimental setup to ensure accurate CO monitoring.
Since the scaling appeared to be animal specific, we divided all 43 experiments according to 4 categories. Categories C1 and C2 exhibited only little amplification of factor α, whereas C3 and C4 showed a strong influence at higher PEEP levels (Fig. 8). It appeared that 80% of the experiments were type C1 (32.5%) and C2 (46.5%).
In the present study, a 16-electrode EIT system was used. Future studies using systems with a higher resolution (e.g., a 32- or 64-electrode system) would allow determination of whether a higher resolution is (or is not) an advantage concerning the possibility of separating the cardiac- and pulmonary-related EIT signals.
Although sample size determination was not performed in the planning phase of our trial, and thus, one could argue that the power was insufficient; the effect to be established was unknown before the trial, and thus, reliable sample size planning could not be performed.
We conclude that EIT is a promising technique for (future) continuous CO monitoring. HR can be determined with a very high level of accuracy. In the present study, the prediction of SV by EIT yielded relatively robust results in 80% of the experiments, but only if the EIT data were calibrated in each dataset according to the reference SVTTD. However, a scaling of the cardiac impedance signal ZCard caused by changes of the cardiac vicinity became apparent in 20% of the experiments, an issue that should be investigated before EIT can be used in clinical practice for purposes of continuous CO monitoring.
Name: Robert Pikkemaat, Dipl.-Ing.
Contribution: This author planned, conducted, and evaluated the study; since this author is an engineer, the medical end points were defined in collaboration with S. Lundin and O. Stenqvist.
Attestation: Robert Pikkemaat approved the final manuscript, reviewed the original study data and the data analysis and is the archival author, but the data are also accessible to Steffen Leonhardt.
Conflicts of Interest: Parts of the study were financed by, and the EIT equipment kindly provided by, Dräger Medical GmbH, Lübeck, Germany.
Name: Stefan Lundin, MD, PhD.
Contribution: This author planned and administered the study partially (i.e., the medical part).
Attestation: Stefan Lundin approved the final manuscript.
Conflicts of Interest: Parts of the study were financed, and the EIT equipment kindly provided, by Dräger Medical GmbH, Lübeck, Germany.
Name: Ola Stenqvist, MD, PhD.
Contribution: This author partially (i.e., the medical part) administered the study.
Attestation: Ola Stenqvist approved the final manuscript.
Conflicts of Interest: This author received a speaker’s honorarium and travel costs from Dräger Medical GmbH, Lübeck, Germany.
Name: Ralf-Dieter Hilgers, Dr. rer. Nat.
Contribution: This author helped in statistical analysis and interpretation of study results.
Attestation: Ralf-Dieter Hilgers approved the final manuscript.
Conflicts of Interest: The author has no conflicts of interest to declare.
Name: Steffen Leonhardt, Dr. med. Dr.-Ing.
Contribution: This author partially (i.e., the technical part) administered the study; this author was also involved in planning the experiments and discussions on analysis of the recorded data.
Attestation: Steffen Leonhardt approved the final manuscript.
Conflicts of Interest: This author discloses financial support for unrestricted research on EIT-based perfusion imaging from Dräger Medical GmbH, Lübeck, Germany. He has also received honoraria for lectures and consultancy.
This manuscript was handled by: Franklin Dexter, MD, PhD.
a In both cases, i.e., correlation of HR and EIT und PEEP as well as under infusion, a multilevel model could not be fitted to the data because an infinite likelihood is assumed in iteration 0 due to a nonpositive definite estimated R matrix. Thus, only crude correlation estimates are presented above. Cited Here...
1. Swan HJ, Ganz W, Forrester J, Marcus H, Diamond G, Chonette D. Catheterization of the heart in man with use of a flow-directed balloon-tipped catheter. N Engl J Med. 1970;283:447–51
2. Mathews L, Singh RK. Cardiac output monitoring. Ann Card Anesth. 2008;11:56–68
3. Connors AF Jr, Speroff T, Dawson NV, Thomas C, Harrell FE Jr, Wagner D, Desbiens N, Goldman L, Wu AW, Califf RM, Fulkerson WJ Jr, Vidaillet H, Broste S, Bellamy P, Lynn J, Knaus WA. The effectiveness of right heart catheterization in the initial care of critically ill patients. SUPPORT Investigators. JAMA. 1996;276:889–97
4. Harvey S, Stevens K, Harrison D, Young D, Brampton W, McCabe C, Singer M, Rowan K. An evaluation of the clinical and cost-effectiveness of pulmonary artery catheters in patient management in intensive care: a systematic review and a randomised controlled trial. Health Technol Assess. 2006;10:iii–iv, ix–xi, 1–133
5. Funk DJ, Moretti EW, Gan TJ. Minimally invasive cardiac output monitoring in the perioperative setting. Anesth Analg. 2009;108:887–97
6. Cholley BP, Payen D. Noninvasive techniques for measurements of cardiac output. Curr Opin Crit Care. 2005;11:424–9
7. Alhashemi JA, Cecconi M, Hofer CK. Cardiac output monitoring: an integrative perspective. Crit Care. 2011;15:214
8. Hadian M, Kim HK, Severyn DA, Pinsky MR. Cross-comparison of cardiac output trending accuracy of LiDCO, PiCCO, FloTrac and pulmonary artery catheters. Crit Care. 2010;14:R212
9. Berberian G, Quinn TA, Vigilance DW, Park DY, Cabreriza SE, Curtis LJ, Spotnitz HM. Validation study of PulseCO system for continuous cardiac output measurement. ASAIO J. 2005;51:37–40
10. Tan HL, Pinder M, Parsons R, Roberts B, van Heerden PV. Clinical evaluation of USCOM ultrasonic cardiac output monitor in cardiac surgical patients in intensive care unit. Br J Anesth. 2005;94:287–91
11. Adler A, Amato MB, Arnold JH, Bayford R, Bodenstein M, Böhm SH, Brown BH, Frerichs I, Stenqvist O, Weiler N, Wolf GK. Whither lung EIT: where are we, where do we want to go and what do we need to get there? Physiol Meas. 2012;33:679–94
12. Leonhardt S, Lachmann B. Electrical impedance tomography: the holy grail of ventilation and perfusion monitoring? Intensive Care Med. 2012;38:1917–29
13. Gabriel C, Peyman A, Grant EH. Electrical conductivity of tissue at frequencies below 1 MHz. Phys Med Biol. 2009;54:4863–78
14. Meier T, Luepschen H, Karsten J, Leibecke T, Grossherr M, Gehring H, Leonhardt S. Assessment of regional lung recruitment and derecruitment during a PEEP trial based on electrical impedance tomography. Intensive Care Med. 2008;34:543–50
15. Nguyen DT, Jin C, Thiagalingam A, McEwan AL. A review on electrical impedance tomography for pulmonary perfusion imaging. Physiol Meas. 2012;33:695–706
16. Barber DC, Brown BH. Applied potential tomography. J. Phys. E: Sci Instrum. 1984;17:723–33
17. Brown BH, Barber DC, Seagar AD. Applied potential tomography: possible clinical applications. Clin Phys Physiol Meas. 1985;6:109–21
18. McArdle FJ, Suggett AJ, Brown BH, Barber DC. An assessment of dynamic images by applied potential tomography for monitoring pulmonary perfusion. Clin Phys Physiol Meas. 1988;9(Suppl A):87–91
19. Eyüboglu BM, Brown BH. Methods of cardiac gating applied potential tomography. Clin Phys Physiol Meas. 1988;9(Suppl A):43–8
20. Fagerberg A, Stenqvist O, Aneman A. Monitoring pulmonary perfusion by electrical impedance tomography: an evaluation in a pig model. Acta Anesthesiol Scand. 2009;53:152–8
21. Fagerberg A, Stenqvist O, Aneman A. Electrical impedance tomography applied to assess matching of pulmonary ventilation and perfusion in a porcine experimental model. Crit Care. 2009;13:R34
22. Smit HJ, Vonk Noordegraaf A, Marcus JT, Boonstra A, de Vries PM, Postmus PE. Determinants of pulmonary perfusion measured by electrical impedance tomography. Eur J Appl Physiol. 2004;92:45–9
23. Smit HJ, Handoko ML, Vonk Noordegraaf A, Faes TJ, Postmus PE, de Vries PM, Boonstra A. Electrical impedance tomography to measure pulmonary perfusion: is the reproducibility high enough for clinical practice? Physiol Meas. 2003;24:491–9
24. Zadehkoochak M, Blott BH, Hames TK, George RF. Pulmonary perfusion and ventricular ejection imaging by frequency domain filtering of EIT (electrical impedance tomography) images. Clin Phys Physiol Meas. 1992;13(Suppl A):191–6
25. Frerichs I, Pulletz S, Elke G, Reifferscheid F, Schadler D, Scholz J, Weiler N. Assessment of changes in distribution of lung perfusion by electrical impedance tomography. Respiration. 2009;77:282–91
26. Krivoshei A, Min M, Parve T, Ronk A. An adaptive filtering system for separation of cardiac and respiratory components of bioimpedance signal. Proceedings International Workshop on Medical Measurements and Applications - MeMeA 2006. 2006 IEEE, Benevento, Italy:10–5
27. Kerrouche N, McLeod CN, Lionheart WR. Time series of EIT chest images using singular value decomposition and Fourier transform. Physiol Meas. 2001;22:147–57
28. Deibele JM, Luepschen H, Leonhardt S. Dynamic separation of pulmonary and cardiac changes in electrical impedance tomography. Physiol Meas. 2008;29:S1–14
29. Solà J, Adler A, Santos A, Tusman G, Sipmann FS, Bohm SH. Non-invasive monitoring of central blood pressure by electrical impedance tomography: first experimental evidence. Med Biol Eng Comput. 2011;49:409–15
30. Borges JB, Suarez-Sipmann F, Bohm SH, Tusman G, Melo A, Maripuu E, Sandström M, Park M, Costa EL, Hedenstierna G, Amato M. Regional lung perfusion estimated by electrical impedance tomography in a piglet model of lung collapse. J Appl Physiol (1985). 2012;112:225–36
31. Kubitz JC, Annecke T, Forkl S, Kemming GI, Kronas N, Goetz AE, Reuter DA. Validation of pulse contour derived stroke volume variation during modifications of cardiac afterload. Br J Anesth. 2007;98:591–7
32. Carlsson M, Cain P, Holmqvist C, Stahlberg F, Lundback S, Arheden H. Total heart volume variation throughout the cardiac cycle in humans. Am J Physiol Heart Circ Physiol. 2004;287:H243–50
33. McKay RG, Spears JR, Aroesty JM, Baim DS, Royal HD, Heller GV, Lincoln W, Salo RW, Braunwald E, Grossman W. Instantaneous measurement of left and right ventricular stroke volume and pressure-volume relationships with an impedance catheter. Circulation. 1984;69:703–10
34. Hamlett A, Ryan L, Wolfinger R. On the use of PROC MIXED to estimate correlation in the presence of repeated measures. SAS Users Group International, Proceedings of the Statistics and Data Analysis Section. 2004;198–29:1–7
35. Bland JM, Altman DG. Statistical notes: Calculating correlation coefficients with repeated observations: Part 2—coefficients between subjects. BMJ. 1995;310:633
36. Roy A. Estimating correlation coefficient between two variables with repeated observations using mixed effects model. Biom J. 2006;48:286–301
37. Hadian M, Kim HK, Severyn DA, Pinsky MR. Cross-comparison of cardiac output trending accuracy of LiDCO, PiCCO, FloTrac and pulmonary artery catheters. Crit Care. 2010;14:R212
38. Costa MG, Della Rocca G, Chiarandini P, Mattelig S, Pompei L, Barriga MS, Reynolds T, Cecconi M, Pietropaoli P. Continuous and intermittent cardiac output measurement in hyperdynamic conditions: pulmonary artery catheter vs. lithium dilution technique. Intensive Care Med. 2008;34:257–63
39. Bajorat J, Hofmockel R, Vagts DA, Janda M, Pohl B, Beck C, Noeldge-Schomburg G. Comparison of invasive and less-invasive techniques of cardiac output measurement under different haemodynamic conditions in a pig model. Eur J Anesthesiol. 2006;23:23–30
40. Patterson RP, Zhang J, Mason LI, Jerosch-Herold M. Variability in the cardiac EIT image as a function of electrode position, lung volume and body position. Physiol Meas. 2001;22:159–66
41. Hofmann D, Klein M, Wegschneider K, Sakka SG. Extended hemodynamic monitoring using transpulmonary thermodilution influence of various factors on the accuracy of the estimation of intrathoracic blood volume and extravascular lung water in critically ill patients (article in German). Anesthesist. 2005;54:319–26