Author + information
- Received August 18, 2004
- Revision received October 5, 2004
- Accepted October 12, 2004
- Published online February 1, 2005.
- James Pemberton, MD*,
- Xiaokui Li, MD*,
- Tara Karamlou, MD†,
- Chloe A. Sandquist, MD†,
- Karl Thiele, PhD‡,
- Irving Shen, MD†,
- Ross M. Ungerleider, MD†,
- Antoinette Kenny, MD§ and
- David J. Sahn, MD, MACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. David J. Sahn, The Clinical Center for Congenital Heart Disease, Oregon Health & Science University, 3181 SW Sam Jackson Park Road, L608, Portland, Oregon, 97239-3098
Objectives The purpose of this study was to investigate whether cardiac output (CO) could be accurately computed from live three-dimensional (3-D) Doppler echocardiographic data in an acute open-chested animal preparation.
Background The accurate measurement of CO is important in both patient management and research. Current methods use invasive pulmonary artery catheters or two-dimensional (2-D) echocardiography or esophageal aortic Doppler measures, with the inherent risks and inaccuracies of these techniques.
Methods Seventeen juvenile, open-chested pigs were studied before undergoing a separate cardiopulmonary bypass procedure. Live 3-D Doppler echocardiography images of the left ventricular outflow tract and aortic valve were obtained by epicardial scanning, using a Philips Medical Systems (Andover, Massachusetts) Sonos 7500 Live 3-D Echo system with a 2.5-MHz probe. Simultaneous CO measurements were obtained from an ultrasonic flow probe placed around the aortic root. Subsequent offline processing using custom software computed the CO from the digital 3-D Doppler DICOM data, and this was compared to the gold standard of the aortic flow probe measurements.
Results One hundred forty-three individual CO measurements were taken from 16 pigs, one being excluded because of severe aortic regurgitation. There was good correlation between the 3-D Doppler and flow probe methods of CO measurement (y = 1.1x − 9.82, R2= 0.93).
Conclusions In this acute animal preparation, live 3-D Doppler echocardiographic data allowed for accurate assessment of CO as compared to the ultrasonic flow probe measurement.
Cardiac output (CO) is an important parameter both in clinical medicine and in experimental biology, providing a guide to the circulatory status and response to therapeutic interventions. Several methods are available for assessment of CO; however, these are not without inherent problems.
Two-dimensional echocardiographic (2-DE) Doppler assessment of CO is derived from the velocity time integral of the pulse wave (PW) Doppler at the level of the aortic valve, multiplied by the valve diameter and heart rate. This 2-DE method has inherent inaccuracies, as errors in diameter measurement are multiplied in the calculations and PW velocity determination may be inaccurate, especially in low-output states (1,2). In addition, the aortic valve orifice diameter is not uniform during the cardiac cycle or with differing hemodynamic states (3,4). There is also the angle dependence of Doppler (1,5), the parabolic flow velocity profile in the aorta (1), and skewing of this profile such that the higher velocities may be found at the septal or posterior side of the outflow tract (6,7), which can be more pronounced in larger animals (8).
Esophageal Doppler monitors can be employed for more long-term CO monitoring in critical care areas. Inaccuracies may be introduced from probe movement, and several assumptions are made about aortic diameters (9,10). This technique is minimally invasive and not suitable for nonsedated patients.
Thermodilution techniques of CO monitoring using invasive pulmonary artery catheters are common in the critical care setting. These techniques are not without errors, especially in the presence of intracardiac or pulmonary shunts, low-output states (11), and tricuspid regurgitation (12). Catheters are invasive and present risks associated with both catheter insertion and long-term use, including vessel perforation, sepsis, and thrombus formation and embolization (13).
Previous work by our group developed a reconstructive three-dimensional echocardiographic (3-DE) technique to accurately compute intracardiac flow volumes and derive CO using transvalvular Doppler data (14–17). Three-dimensional echocardiographic methods of flow volume computation overcome the angle dependence of 2-D Doppler (18) and the limitations of standard 2-DE methods of volume computation. However, these studies were limited by the time-consuming acquisitions that required multiple 2-DE slices to be acquired and reconstructed into a 3-D image. The aim of this study was to assess the accuracy of a new implementation of live 3-D Doppler echocardiography in the computation of CO in an open-chested animal. With this new technology allowing the Doppler data to be acquired as a complete 3-D volume over only seven alternate gated cardiac cycles, the need for lengthy acquisition times or probe position changes is abolished.
Seventeen juvenile Yorkshire pigs weighing 2.7 to 11.3 kg (mean 4.29 kg) were studied before undergoing a separate cardiopulmonary bypass (CPB) procedure. All animals were intubated and ventilated for the duration of the procedure, and anesthesia was maintained with 2% isofluorane and oxygen. The Institutional Animal Care and Use Committee, Oregon Health and Science University approved all procedures and operations.
Live 3-D Doppler echocardiography
Live 3-D Doppler images were acquired using the Philips Medical Systems (Andover, Massachusetts) Sonos 7500 Live 3-D Echocardiography system with a 3000-element array, 2.5-MHz xMATRIX transthoracic probe. For 3-D color Doppler acquisitions, ensuring adequate frame rate and packet size and capture of the full cardiac cycle, the system obtains a 30° × 30° color Doppler pyramid over 7 alternate gated cardiac cycles (14 cycles in total). The resulting dynamic 3-D tissue and color Doppler volume can be reviewed and navigated through immediately to ensure all areas of interest have been captured.
Apical epicardial scanning was performed to obtain 3-D color Doppler volumes incorporating the left ventricular outflow tract (LVOT), aortic valve, and ascending aorta with electrocardiographic gating. Depth settings were adjusted as appropriate in each animal (approximately 6 cm) and the Nyquist limit set to approximately 100 cm/s to avoid aliasing and to maximize packet size and frame rate (15 to 20 Hz). An ultrasound gel-filled probe cover (Advance Medical Designs Inc., Atlanta, Georgia) was used over the probe, and ultrasound gel was used epicardially for optimal contact with the heart.
Between 2 and 16 3-D image loops were obtained from each animal under stable hemodynamic conditions. Loops were viewed after acquisition to ensure that adequate data was obtained; specifically, that the whole of the LVOT and aortic valve were contained in the dataset (Fig. 1A).
Ultrasonic flow probes
A 10-mm ultrasound flow probe and blood flow meter (Transonic Systems, Inc., Ithaca, New York) was used to accurately measure CO in each animal. The probe was placed around the aortic root just distal to the origin of the coronary arteries in 11 animals and around the pulmonary root in the remaining animals. A CO value (ml/min) was recorded for each 3-D acquisition in all animals. Probes were calibrated to zero and maximal flow before each study. The flow probes give mean flow rate on a beat-to-beat basis.
After completion of the study, 3-D Doppler data was transferred to a personal computer for flow volume computation using specially designed computer software (4D Echo-View, TomTec Imaging Systems, Unterschleissheim, Germany). To achieve this, the software uploads the full 3-D volume dataset in proprietary 3-D Cartesian DICOM format. Controls allow for the cut plane of the 3-D volume to be moved in two perpendicular longitudinal B-planes to ensure that the LVOT is clearly seen on both cut surfaces. A sampling curve, the curved surface of a hemisphere, is then placed on these perpendicular B-plane images at the level of the aortic valve (Fig. 1B). An arrow is used to indicate the direction of flow to ensure the software accurately knows this.
The software then projects the velocity data for each frame of the cardiac cycle to give a color image of the LVOT at the level of the sampling curve, as if viewed from the left ventricular cavity perpendicular to the aortic valve. A sampling area is then traced around the region of interest—in this case, the aortic outflow (Fig. 1C). The software automatically reads the retained digital velocity assignments in the dataset, within the region of interest, and then computes the flow velocity profile and stroke volume for that given area over the cardiac cycle from the velocity data (Fig. 1D). In the absence of significant aortic regurgitation, the aortic forward flow volume should equal the stroke volume. The CO is calculated by multiplication of the aortic forward flow with the heart rate. The heart rate is stored as part of the 3-D Doppler dataset on the echo system.
All values are expressed as mean ± SD. Linear regression analysis was used to obtain correlation coefficients between the 3-D Doppler and reference flow probe data. Mean bias agreement between the two methods was tested according to the Bland-Altman method with a 95% confidence interval (19).
Ten randomly selected datasets were analyzed by a second operator (X. L.) who was blinded to the results of both the first operator and the reference flow probe data.
One animal was excluded because of severe aortic regurgitation; there were no other significant valvular abnormalities in any of the other animals. All were in sinus rhythm and studied under stable baseline hemodynamic conditions. In addition, four animals were also studied after CPB under a different but stable hemodynamic state. Heart rates in the animals ranged from 75.6 to 150 beats/min (mean: 112.8 ± 17.7 beats/min) and were stable for the duration of each study period.
One hundred forty-three individual CO measures with corresponding flow probe reference values were analyzed from the 20 separate studies (7.15 observations per study). There was good correlation between CO derived from 3-D Doppler data and that of the reference flow probe: y = 1.1x − 9.82, R2= 0.93 (Fig. 2).
When the 3-D Doppler and flow probe data for each study were averaged and compared, as is the case in clinical practice, where an average of several CO measurements would be used rather than a single point reading, there was excellent agreement between the two methods: y = 1.11x − 17.5, R2= 0.99 (Fig. 3).
Analysis of the level of agreement between the two methods in the individual data shows a mean difference of 30.6 ml/min and a standard deviation of 70 ml/min, with 3-D Doppler data tending to overestimate CO compared to the reference flow probe (Fig. 4).Analysis of the combined data for each animal shows a mean difference of 31.65 ml/min and a standard deviation of 46 ml/min.
The interobserver variability was tested using 10 randomly selected datasets by a second observer (X. L.). There was good correlation between the results of the two observers (y = 1.28x − 72.7, R2= 0.96). Mean difference was 62.6 ml/min with a standard deviation of 77 ml/min.
Real-time 3-DE technology has previously been utilized to assess stroke volume and, hence, CO by ventricular volume measurement. Several studies have investigated this against reference stroke volume calculation and shown it to be an accurate method of CO estimation both in vitro (20,21) and in vivo (22,23). However, time-consuming offline processing is required and border detection can still be a source of error. One previous study (24) used an earlier real-time 3-D Doppler system to compute flow at the level of the LVOT. In that study, the low frame rate of the Doppler acquisition, approximately 10 Hz, led to underestimation of flow volume. Accuracy was improved by development of a mathematical correction.
The results of this study show that live 3-D Doppler data can be used to accurately compute CO from the 3-D transaortic Doppler profile, as in this animal model. The resulting increased voxel density from the high line density, frame rate, and relatively shallow depth lead to improvement in temporal resolution and, therefore, Doppler data for flow volume estimation. There is improved spatial resolution with the use of the 3000-element array 4D probe on the Philips Sonos 7500 system.
Cardiac output values in these animals were in the physiological ranges for animals of this size, approximately 100 ml/kg (25,26). In this study, the 3-DE results overestimated the CO by 7.3%. The flow probe was placed distal to the coronary artery ostia, whereas the 3-DE estimations were made at the level of the aortic valve. The flow probe measures mean aortic flow on a beat-to-beat basis, whereas the 3-DE method measures systolic forward flow only. The difference between the two methods could be accounted for by coronary blood flow, a large proportion of which is diastolic. We did not attempt to measure coronary artery flow in this study.
The 3-DE overcomes some of the problems seen with 2-DE, including the assumption of a circular aortic valve orifice (3–5) and the flat flow velocity profile over the aortic valve (7,8). This live 3-D Doppler technique overcomes these limitations by integration of all the flow velocity vectors in the sampled area over a frame-by-frame time period to yield a transaortic flow volume.
Previous work by our group used reconstructive 3-DE data to derive intracardiac flow volumes (14–18). This has been shown to be accurate, but had some limitations, including time needed for data acquisition and processing, and data mismatch due to movement or respiration artifact. This advancement in 3-D technology, allowing acquisition of 3-D volumes over seven alternate gated cardiac cycles and displayed as a full 3-D color Doppler volume, reduces the problem of variations in measurements due to time taken for acquisitions and the need for lengthy patient cooperation/sedation or the need for transesophageal echo. It also overcomes the cumbersome nature of 3-D gating with the Flock of Birds or stepper motors systems.
Despite the rapidity of acquisitions, CO computation still takes place offline. This does not allow for instantaneous CO monitoring. However, offline processing of the data is quick and the learning curve for the software, short. The additional advantage of the use of this echo technique is the extra information gained regarding cardiac structure and function.
In this study, the presence of significant aortic regurgitation was an exclusion criterion, as the flow probe only gave mean aortic flow on a beat-to-beat basis. Turbulent flow and flow velocities above the Nyquist limit for the system may limit accurate flow-volume calculation. In this study, we did not encounter problems with aliasing in the LVOT at the Nyquist limit used. The analysis program allows for “baseline shift.” This option allows for some of the aliased velocities to be incorporated into the flow velocity profile; we did not use the baseline shift option for our analysis.
As the 3-D acquisition is over seven alternate cardiac cycles, stable hemodynamic conditions are needed over this time period. In patients with higher heart rates, the acquisition will be faster than with slower heart rates. In our study, with an average heart rate of 112 beats/min, acquisition times were approximately 7.5 s, and hemodynamic conditions in our animals were stable over this short time period. The maximum frame rate for 3-D Doppler acquisition is 20 Hz, and in our studies with higher heart rates, this was achieved.
A regular cardiac rhythm is also necessary, as irregularities such as atrial fibrillation would produce volume mismatch over the seven-gated sub-volumes. Irregular rhythms are a problem with most methods of cardiac output determination as the stroke volume is different with each beat depending on filling time (27). This could be overcome by averaging a number of CO estimates, as is often the case clinically in atrial fibrillation (28).
The animals used in this study were small (average 4.29 kg) but could serve as a model for neonate echocardiography. In these animal preparations, epicardial scanning in a controlled experimental environment was used. This allowed optimal 3-D Doppler data quality acquisition. In addition, depth settings were approximately 6 cm, allowing for excellent imaging of the LVOT and aortic valve. In clinical applications, depth settings, especially in adult patients, may be such that the frame rate and Nyquist limit would not allow for adequate velocity data for accurate flow volume computation. Further studies will be required to clarify the maximal depth settings possible for flow volume computation.
The 3-D probe currently uses a 2-to-4 MHz range for scanning. In adult patients with poor echo windows, it can be used with tissue harmonics to improve image quality. In pediatric patients, this, or a higher frequency, would be preferable for improved resolution.
In this study we used an ultrasonic flow probe as the reference for CO calculation. We felt that this was the most accurate method of CO calculation in these animals given the small size of the vessels (<10 mm) and the potential for inaccurate diameter and PW measures. The extension of this study to involve both adult and pediatric human subjects in a clinical setting is the next step in defining the clinical use of live 3-D Doppler technology in the computation of flow volumes. In addition, it will be possible to assess the difference against standard 2-DE methods. Our initial findings in human studies suggest good correlation from 3-DE against reference volume calculation. It may also be possible to extend this technology to quantify flow over cardiac valves and, therefore, to assess valvular insufficiency as a product of forward minus reverse flow volumes.
This study has shown that live 3-D Doppler echocardiography can be successfully used to compute intracardiac flow volumes accurately in this animal model and holds promise for a noninvasive technique for flow-volume assessment in human subjects in the future.
Dr. Thiele is an employee of Philips Medical Systems, and Dr. Sahn is an occasional consultant to the company. Dr. Roberto Lang acted as Guest Editor.
- Abbreviations and acronyms
- aortic valve
- cardiac output
- cardiopulmonary bypass
- left ventricular outflow tract
- pulse wave
- two-dimensional echocardiography
- three-dimensional echocardiography
- Received August 18, 2004.
- Revision received October 5, 2004.
- Accepted October 12, 2004.
- American College of Cardiology Foundation
- Marshall S.A.,
- Weyman A.E.
- Sahn D.J.,
- Valdes-Cruz L.M.
- Quinones M.A.,
- Otto C.M.,
- Stoddard M.,
- Waggoner A.,
- Zoghbi W.A.
- Matre K.,
- Kvitting P.,
- Zhou Y.Q.,
- Faerestrand S.
- Mori Y.,
- Rusk R.A.,
- Jones M.,
- et al.
- Schindera S.T.,
- Mehwald P.S.,
- Sahn D.J.,
- Kececioglu D.
- Qin J.X.,
- Jones M.,
- Shiota T.,
- et al.