Automated Cardiac Output Measurement by Spatiotemporal Integration of Color Doppler Data
In Vitro and Clinical Validation
Background A new Doppler echocardiographic technique has been developed for automated cardiac output measurement (ACOM) that assumes neither a flat flow profile nor collinearity with the scan line, but clinical validation of this method is lacking.
Methods and Results In 165 subjects (50 intensive care patients, 10 dobutamine echocardiography patients, and 105 normal volunteers; age, 49.4±19.3 years; 92 men), ACOM was performed in the left ventricular outflow tract (LVOT), with the color baseline shifted to avoid aliasing. ACOM was also tested in a pulsatile in vitro model. Stroke volume was calculated by double integration of Doppler signals in space (across the LVOT) and in time (through the systolic period), assuming hemiaxial symmetry: ∫∫π r v(r,t) dr dt, where v(r,t) is the velocity at a distance r from the center of the LVOT at time t during systole. Stroke volume from ACOM was compared with thermodilution (TD), aortic valve pulsed-wave Doppler (PWAO), and left ventricular echocardiographic (two-dimensional [2D]) methods. There was good correlation between ACOM and PWAO (r=.93), TD (r=.86), and 2D (r=.74), with close agreement seen. ACOM had higher correlation and agreement with TD than did either PWAO (P<.02) or 2D (P<.01). ACOM was also able to track accurately the changes in cardiac output with dobutamine infusion in comparison with PWAO (r=.94). In vitro assessment demonstrated excellent correlation (r=.98, y=1.0x+1.94) with little impact of pulse repetition frequency or misalignment up to 30°. Gain dependency was noted but could be optimized by visual inspection of the color image.
Conclusions Automatic integration of numerical data within color Doppler flow fields is a feasible new method for quantifying flow. It is simpler and faster, requires fewer assumptions, and uses only one apical view. ACOM is a promising new approach to echocardiographic quantification that deserves further study and refinement.
Accurate measurement of cardiac output is important in assessing seriously ill patients and monitoring the impact of therapeutic maneuvers. The Fick oxygen method is generally considered the most accurate technique available for measuring cardiac output,1 but it is time consuming and requires an expert operator to achieve optimal performance. The TD method is easier to perform2 3 than the Fick method, but the need for catheterization of the right side of the heart has limited its applicability to intensive care and intraoperative settings.
The Doppler measurement of blood flow velocities in human aorta was first described by Light4 in 1969, and a variety of echocardiographic methods have been reported to quantify flow in the heart and great vessels, most commonly the LVOT.5 6 7 8 9 10 Although this noninvasive method has been reported to be accurate in carefully controlled studies, it requires two echocardiographic imaging views and meticulous care in measuring the area of the LVOT and the time velocity integral of systolic velocity. Alternatively, ventricular stroke volume can be calculated as the difference between end-systolic and end-diastolic volumes, obtained from 2D echocardiograms with use of hand-traced10 11 or automatically detected endocardial boundaries,12 13 but this also requires careful technique and may yield inaccurate results.
A new technique has recently been developed for ACOM, which uses digital velocities from a color Doppler flow map to perform spatiotemporal integration across the LVOT throughout systole.14 Although promising in vitro results have been reported, clinical data are scarce. The purpose of this study, therefore, was to evaluate the accuracy of this new method of ACOM in a large group of patients and normal subjects.
The initial study population consisted of 155 subjects with a mean age of 49.3±19.3 years (range, 18 to 85 years). One group included 60 critically ill patients in medical intensive care units of The Cleveland Clinic Foundation with in-dwelling Swan-Ganz catheters. Ten patients were excluded from this study because of more than mild tricuspid regurgitation (n=3), more than mild mitral regurgitation (n=3), and inadequate echocardiographic images (n=4 with intra-aortic balloon pumps who could not turn on their sides). The 50 remaining patients (66.6±9.8 years; range, 27 to 84 years; 29 men) had diagnoses of unstable angina (n=18), acute myocardial infarction (n=8), dilated cardiomyopathy (n=6), respiratory failure (n=8), heart failure (n=8), and amyloid heart disease (n=2). Of the 50 patients, 8 were mechanically ventilated. In this group, cardiac output was measured by four methods, detailed below: ACOM, manual tracing of end-diastolic and end-systolic 2D echocardiographic volumes (the 2D method), PWAO, and TD.
Another group of subjects included 105 volunteers ranging in age from 18 to 67 years (mean±SD, 38.4±13.3 years; 58 men). In these subjects, cardiac output was measured by the methods above except for TD.
Finally, in 10 additional patients (70.6±13.3 years, 5 men) undergoing dobutamine echocardiographic examinations, cardiac output was determined by ACOM and PWAO at baseline, peak dobutamine dosing, and several intermediate doses (33 total stages).
Cardiac Output Techniques
Automated Cardiac Output Measurement
Echocardiography was performed with a commercially available machine (Toshiba SSA-380A) equipped with prototype software for automated cardiac output measurement. Images were obtained from the apical long-axis view with a 2.5-MHz probe and oriented so that the LVOT was optimally visualized. The color baseline was shifted so that unaliased (blue) flow was seen everywhere in the LVOT throughout systole. Scan line density was adjusted to achieve a frame rate of 28 to 35 frames per second. A region of interest was placed within the color sector across the LVOT at the level of the mitral annulus (Fig 1⇓). When optimized, color images were stored in a 256-image buffer for further analysis. On a selected beat, the systolic period was manually defined by a trigger mark based on the ECG and the appearance and disappearance of blue color flow in the LVOT. Once approved, stroke volume was automatically calculated by double integration of Doppler signals in space (across the full width of the LVOT) and in time (frame by frame throughout the systolic period), assuming hemiaxial symmetry: ∫∫π r v(r,t)dr dt, where v(r,t) is the velocity at a distance r from the center of the LVOT (defined on each frame as the centroid of the Doppler velocity profile) and at time t during systole. Cardiac output was then calculated by multiplying stroke volume by heart rate, obtained automatically from the ECG. The measurements from three to five beats were averaged. This technique was performed by an investigator blinded to the results of the other three methods for measuring cardiac output. In 10 normal volunteers, ACOM measurements were repeated from the apical five-chamber view.
Pulsed Doppler Technique
The PWAO and 2D methods have been detailed previously and thus are described briefly here. Pulsed Doppler echocardiography was performed by positioning the sample volume within the aortic valve annulus as visualized from the apical long-axis view. The aortic annulus diameter D (from the anterior to posterior aortic leaflet hinge points) was obtained from the parasternal long-axis view, and area A was calculated as πD2/4. The time-velocity integral of aortic annular flow was obtained by manual tracing of the pulsed Doppler recording and multiplied by area A to yield stroke volume and subsequently by heart rate to obtain cardiac output. The average of three to five consecutive beats was used.
Two-Dimensional Echocardiographic Technique
Left ventricular volumes were obtained from manual endocardial traces at end diastole and end systole imaged from the apical four-chamber view with volume calculated by use of the single plane area-length formula. Both the PWAO and 2D calculations were performed by an investigator without knowledge of automated cardiac output measurements and results of TD. For logistical reasons, 2D volumes were not measured in all subjects but were obtained in 42 patients and 78 volunteers (n=120). Typically, this was due to the patient having another clinical or procedural appointment that limited the time he or she was available for the study. By prospective design, we declared the PWAO and TD measurements to be of the highest priority to compare with ACOM.
A Swan-Ganz TD catheter was inserted into the pulmonary artery in all study patients for clinical indications. Cardiac output was measured by injection of 10 mL of 0.9% saline solution at room temperature in the right atrium. The measurement was repeated three to five times until three values were within 10% of each other. TD measurements of cardiac output were done by an intensive care unit nurse who was blinded to the results of the other three methods and were performed within 10 minutes of the ACOM calculation with no significant intervening change in clinical status noted.
In Vitro Study
To further test the accuracy of flow rate calculation by ACOM, we examined flow in an in vitro model with carefully controlled hydrodynamics. The EchoCal model CD10 (Dynatek Laboratories, Inc) was used to produce sinusoidal pulsatile flow with a flat velocity profile across a 1.0-cm tube. Peak velocity ranged from 40 to 100 cm/s; peak flow rate ranged from 33 to 79 mL/s. A 2.5-MHz echocardiographic probe (Toshiba SSA-380A) was used to record color and spectral pulsed Doppler data, with the ultrasound beam parallel to the direction of flow. Stroke volume was calculated by ACOM by baseline shifting the color data until no aliasing occurred during a selected half pulse (forward or reverse) and manually bracketing the flow temporally for automated spatiotemporal integration. Because of the known sinusoidal shape of the EchoCal pulse [Q(t)=Qmax sin 2πt/td, where Q(t) and Qmax are instantaneous and peak flow rate, respectively, and td is the period of a complete cycle], Qmax could be obtained from stroke volume (SV) as Qmax=πSV/td. Peak flow and stroke volume were also obtained by multiplying the tube area (0.785 cm2) by the peak and time integrals, respectively, of pulsed Doppler velocity.
Impact of Technical Factors on ACOM Accuracy
The impact of receiver gain on ACOM accuracy was assessed both in the normal volunteers and in vitro. In 10 volunteers, pulsed Doppler measurements of ventricular stroke volume and cardiac output were obtained, along with ACOM measurements. ACOM measurements were initially obtained with the color gain optimized so that the color Doppler display of flow filled the LVOT but did not bleed outside this area. These measurements were then repeated with gain systematically turned down approximately one-quarter turn to where some color dropout was apparent in the LVOT and again with the gain approximately one-quarter turn above optimal to where color was seen outside the lumen of the LVOT.
Similar comparisons were obtained in a steady-flow in vitro model. Water with 1% suspended cornstarch was directed through excised bovine aortas (n=4; diameter, 2.0 to 2.5 cm) at flow rates ranging from 93 to 260 mL/s. Gain was set to four levels: (1) obvious color dropout within the aortic lumen, (2) color seen throughout but not outside the lumen, (3) color seen 1 to 2 mm outside the lumen, and (4) color seen >2 mm outside the lumen. Flow was measured by ACOM and expressed as a ratio of true flow (obtained by timed collection). To assess the impact of beam angulation on accuracy, ACOM measurements were repeated with the ultrasound transducer oriented 10°, 20°, 30°, 40°, and 50° from the direction of blood flow through the ex vivo aorta. Similarly, the impact of aliasing velocity was assessed by varying the pulse repetition frequency from 3000 to 8000 Hz in 1000-Hz increments. In each case, flow was expressed as the fraction of true flow, with linear regression used to assess the impact of the varied parameter.
Interobserver and Intraobserver Variabilities
Two investigators independently performed the automated cardiac output measurement on 10 randomly selected subjects within 30 minutes of each other. One investigator obtained repeated ACOM measurements on 10 patients. The mean and standard deviation of the difference of these measurements are reported both in absolute terms and normalized to the average of the two measurements.
All values are expressed as mean±SD. Least-squares linear regression analysis was used to correlate stroke volume and cardiac output derived by ACOM with the TD, PWAO, and 2D methods. In addition, the mean and SD of the difference of two methods (with the reference measurement subtracted from the ACOM measurement) are reported both in absolute terms and normalized to the average of the two measurements.15 Additionally, the difference in the two methods is plotted against their means to better reflect variability in the measurement as a function of measurement magnitude. For the dobutamine echocardiography patients, cardiac output was assessed by ACOM and PWAO. These were compared by linear regression and analysis of agreement for both actual cardiac output and the change from the baseline value. A value of P<.05 was considered statistically significant. Correlation coefficients were compared following Fisher's z transformation. The agreement of three test methods (ACOM, PWAO, and 2D) with the TD reference method was assessed by the F test for equal variances, and paired t testing was used to assess the squared error in the methods.
Feasibility of ACOM
Of the 60 intensive care patients considered for this study, 3 were excluded because of moderate to severe mitral regurgitation because the 2D stroke volume would not be comparable with the other methods, and 3 were excluded for moderate to severe tricuspid regurgitation because of known errors with TD. Cardiac output was successfully determined by ACOM in 50 of the remaining 54 patients (92.6%), with 4 patients having technically inadequate images for ACOM. ACOM was feasible in all 105 volunteers (100%) and the 10 dobutamine patients.
Correlation and Agreement Between Cardiac Output Methods
Similar mean values were observed for stroke volume and cardiac output when measured by the four methods: 53.7±14.0 mL and 4.5±1.0 L/min for ACOM, 53.9±14.9 mL and 4.6±1.0 L/min for PWAO, 53.3±15.6 mL and 4.6±1.1 L/min for 2D, and 56.6±16.4 mL and 4.9±1.2 L/min for TD, respectively (P>.05 by ANOVA). Linear regression analysis demonstrated highly significant correlations between the stroke volume determined by ACOM versus PWAO, TD, and 2D methods (Figs 2 through 4⇓⇓⇓). Some variation was observed, however, as shown in Table 1⇓, with the best correlation for cardiac output noted between ACOM and PWAO (r=.93, P<.0001) and the weakest noted between ACOM and 2D volumes (r=.74, P<.001). Significantly, the correlation between stroke volume measured by TD and ACOM was better than that noted between TD and PWAO (P=.07) and between TD and 2D (P=.02), with the correlations compared following Fisher's z transformation.
Agreement Between Cardiac Output Methods
The agreement between stroke volume determined by ACOM and PWAO was excellent, with a 95% confidence interval of −8.7 to 11.3 mL. The normalized difference between cardiac output determined by ACOM and PWAO was −2.4±8.7% (range, −24.6% to 26.5%); between ACOM and TD, −4.6±13.6% (range, −37.2% to 25.5%); and between ACOM and 2D volumes, −1.9±19.6% (range, −42% to 41%). The percent difference between cardiac output measured by PWAO and TD showed a larger variance and absolute error (P<.02 compared with the difference of ACOM and TD) with −2.5±16.5% (range, −32.8% to 41%), with still larger error noted between TD and 2D volumes (Δ=7.7±19.6%; range, −47% to 43%; P<.01 compared with ACOM accuracy).
Accuracy of ACOM With Varying Cardiac Output
Fig 5⇓ demonstrates the ability of ACOM to track cardiac output in patients undergoing dobutamine echocardiography. For 33 stages in 10 patients, ACOM (y) was quite accurate compared with PWAO measurements (x): y=1.00×−0.04; r=.95, P<.0001; ΔCO(y−x)=−0.04±0.54 L/min. When the change in cardiac output from the baseline measurement (with the increase ranging from 0.1 to 6.6 L/min) is considered, ACOM (y) still demonstrated accuracy: y=1.02×−0.22; r=.94, P<.0001; Δ(ΔCO)=−0.2±0.6 L/min.
Reproducibility of Results
The relative and absolute interobserver variabilities for stroke volume by ACOM were 0.5±5.0% (absolute range, 0.6% to 10.5%) and 0.8±4.8 mL (range, 0.4 to 10.9 mL), respectively, with excellent correlation (r=.98; y=1.0x+0.15; P<.0001). Intraobserver measurements similarly showed acceptable variability (0.1±6.5% and 0.1±3.9 mL; absolute ranges, 0.6% to 10.7% and 0.6 to 7.2 mL) and correlation (r=.98; y=0.94x+0.39; P<.0001).
In Vitro Study
Good agreement between ACOM (31.2±1.38 mL) and PWAO (31.3±0.69 mL) estimates of stroke volume was found (Δstroke volume=−0.1±1.2 mL). Peak flow rate calculated by ACOM (y) was highly correlated with peak PWAO measurements (x): (r=.98; y=1.0x+1.94; Fig 6⇓).
Impact of Technical Factors on ACOM Accuracy
Impact of Imaging Window
In 10 volunteers in whom ACOM measurements were obtained from both the apical long-axis and apical five-chamber windows, there was good correlation (r=.88) and excellent agreement (Δ=1.9±4.1 mL) for stroke volume measurements. There was no significant difference in the agreement of these two data sets with the reference standard PWAO.
Impact of Receiver Gain
In 10 volunteers, significant and predictable variations in ACOM calculations were observed when receiver gain was varied over a wide range. When gain was optimized (color seen throughout the LVOT but not in the wall), cardiac output was 3.9±0.5 L/min. With gain reduced so that color dropout was clearly seen within the LVOT, calculated cardiac output was 0.8±0.5 L/min less than with optimal gain (P=.001), whereas increasing gain to produce artifactual color outside the LVOT caused cardiac output to be 0.7±0.6 L/min more than the baseline measurement (P=.007).
Significant changes in gain also altered ACOM calculations in vitro. Optimal gain led to measured flow that was 98±3% of true flow, while too low a gain (visible color narrower than aortic lumen) caused measured flow to be 69±21% of true flow. Excessive gain led to flow overestimation: 111±8% of true flow for visible color 1 to 2 mm wider than the lumen and 144±25% for color >2 mm wider than lumen (P<.0001 for linear trend with changing gain). In both the clinical and in vitro settings, it was quite evident to the examiner when color gain was either insufficient or excessive.
Impact of Interrogation Angle
Table 2⇓ demonstrates the impact of interrogation angle on flow accuracy. For angles ≤30°, measured flows averaged within 5% of true flow. With increasing skewness, flow underestimation occurred, reaching about 30% for 50° misalignment.
Impact of Pulse Repetition Frequency
After adjusting for the impact of gain and imaging angle, we could demonstrate a barely significant impact of pulse repetition frequency on flow rate accuracy (P=.05). This appeared to be due mainly to underestimation with the combination of high pulse repetition frequency (8 kHz) and low flow rates when the lumen velocity was low (less than half) relative to the Nyquist frequency.
The present study demonstrates that automated analysis of color Doppler velocities in the LVOT can accurately estimate cardiac output in routine clinical practice. This technique was shown to be feasible in 93% of intensive care patients, even those with mechanical ventilation; all of the 105 normal ambulatory volunteers; and 10 dobutamine echo patients. The correlation between cardiac output determined by ACOM and PWAO was excellent, and the correlation between TD and ACOM was superior than that between TD and either PWAO or 2D. We further showed the capability of this technology to track changes in cardiac output in patients undergoing dobutamine echocardiography. These data indicate that automated spatiotemporal integration of color Doppler velocities could be an accurate and practical noninvasive method for determining cardiac output.
Contemporary echocardiographic instruments produce color velocity maps by the autocorrelation of successive ultrasound pulses.16 Although incapable of yielding velocity spectra (such as that obtained by Fourier transformation in standard pulsed or continuous-wave Doppler), the mean velocity estimate obtained by autocorrelation has been shown to accurately reflect true velocity in nonturbulent flow. Although quantitative analyses of digitally output color Doppler data have been reported,17 18 inclusion of such analysis algorithms into commercial products has been rare.
The ACOM algorithm estimates cardiac stroke volume by integrating velocities across the LVOT within each color Doppler frame (assuming hemiaxial symmetry) and then integrating the data from all frames in systole. Importantly, it is not necessary to estimate LVOT size from the B-mode image; rather, the color extent alone is used, explaining in part the gain dependence of the measurements. Furthermore, ACOM exploits the fortuitous partial cancellation of two factors that separately would cause overestimation or underestimation of flow. The first factor is the angle dependence of Doppler. For a Doppler beam misaligned with the flow direction by an angle 𝛉, velocity is underestimated by cos 𝛉; conversely, if the region of interest is misaligned with the LVOT by an angle 𝛉, the LVOT will appear wider than it actually is by 1/cos 𝛉, roughly canceling out the Doppler underestimation of velocity (some error will remain from the out-of-plane measurement of LVOT size). Thus, as long as flow is parallel to its vessel—generally true for the LVOT—then the two factors should largely offset each other. Although not tested clinically in this study, our in vitro data demonstrate that more extreme misalignment (>30°) may cause flow underestimation and may reduce the signal-to-noise ratio unacceptably.
Comparison With Other Methodologies
Pulsed Doppler Echocardiography of Aortic Valve Flow
A variety of studies5 6 7 8 9 19 20 21 22 23 24 have indicated the feasibility and accuracy of determining cardiac output by use of transthoracic Doppler echocardiography of aortic flow with correlations with Fick or TD methodology ranging from .84 to .97. In the present study, a comparable correlation was observed between ACOM and PWAO estimates of cardiac output, although the correlation with TD and ventricular volumetric methods was slightly lower.
Despite its proven accuracy, the PWAO assessment of aortic flow has unfortunately seen only limited use in routine clinical practice because careful recordings must be obtained from two imaging windows; errors in any measurement (particularly the diameter of the aortic annulus because this is squared in the calculations) will be propagated into the final estimate. With ACOM, it is not necessary to measure the aortic diameter accurately, and misalignment of the Doppler beam to the blood flow is largely canceled in the calculations.
Ventricular Volumetric Methods
In the absence of mitral or aortic regurgitation, the difference between ventricular end-diastolic and end-systolic volumes equals the forward stroke volume, and previous studies have validated the utility of echocardiography in this approach.10 11 Unfortunately, subtraction amplifies the relative error of two measurements, requiring meticulous tracing of the endocardium from high quality from 2D images.11 The technique is further limited by the geometric assumptions necessary for calculating volumes. These limitations were evident in the current study because the 2D method consistently had the lowest correlation with whichever reference standard was chosen.
Automated boundary detection has made volumetric quantification of ventricular stroke volume feasible in most patients,12 but this method also is highly operator dependent, requiring extensive practice and patience to obtain reliable data. Suboptimal images from patients with obesity, lung disease, or mechanical ventilation further limit the accuracy of automated boundary detection.
One distinct advantage of the ACOM method is its ease of application. It uses color Doppler information from a single echocardiographic window (typically the apical long-axis or five-chamber view), so it is easily integrated into a clinical examination without the need for storing interim results. We found that optimizing the image, positioning the region of interest, and performing the calculation added about 20 seconds to the examination (of which 5 to 10 seconds was for the automated integration itself). In contrast, the pulsed Doppler method, even under ideal conditions, added more than three times as much time (65 seconds) to the examination, with the 2D volumetric method taking even longer. There is, of course, a learning curve with ACOM, but we have found that experienced sonographers (already capable of acquiring high-quality apical images) could obtain reliable results within about 10 cases. For the present study, the lead author (Dr Sun) performed 10 studies before beginning to assess the normal volunteers. The subsequent 10 patients had high correlation (r=.93) with PWAO measurements but with some underestimation (Δstroke volume=−5.3±3.8 mL). For the final 10 of the 105 volunteers, the correlation was improved (r=.97), as was the degree of underestimation (Δstroke volume=−1.9±4.0 mL), suggesting a small, continued training effect.
In Vitro Study
The results of the in vitro study provide an excellent foundation for the clinical study, demonstrating a close agreement between ACOM calculation and the known stroke volume, with excellent correlation to the measured peak flow rate. It provides significant corroboration with earlier, more comprehensive in vitro experimentation.14
Our in vitro and clinical data demonstrated clear gain dependency of ACOM flow calculations. Importantly, however, it was readily apparent to the examiner when flow rate would be underestimated (areas of color dropout seen within the LVOT) or overestimated (color seen in the wall or outside the LVOT). Thus, although care is needed in adjusting gain, abundant visual clues guide the examiner to the appropriate value. The only significant effect of pulse repetition frequency was seen in vitro with the combination of high Nyquist velocity and low flow velocity, in areas where relatively few color bins were occupied and the limited low-velocity resolution led to flow underestimation. Although similar results were obtained with both the apical long-axis and five-chamber windows, optimization of the alignment seemed to be slightly easier with the long-axis view.
Overall image quality clearly affects the accuracy of the ACOM method, and four patients were excluded from the intensive care unit part of the study because of unusable images. Although the remainder of the images were of high enough quality to perform ACOM, it appeared that the patients who were intubated generally had somewhat poorer correlation with TD.
In the clinical study, no true “gold standard” was available for comparison, although TD, PWAO, and echocardiographic methods have all been used as reference standards in previous studies. ACOM assumes that the velocity recorded at a certain point across the LVOT represents velocities in a hemicircular arc on that side and at that distance from the center of the LVOT. Prior quantitative analysis of velocity profiles in the LVOT confirms the validity of this assumption in most patients.18 Marked distortions in LVOT geometry such as hypertrophic cardiomyopathy certainly will affect this, but such patients were not encountered in this study. However, we have subsequently seen one example of this outside the study: in an elderly woman with a marked upper septal bulge, the LVOT was not axisymmetric, and ACOM measurement from the apical long-axis window predicted a stroke volume of 36 mL, while pulsed Doppler assessment of aortic annular flow measured 73 mL. Furthermore, it is possible that an anteriorly directed jet of mitral regurgitation might flow adjacent to the LVOT and thus be included in the flow estimation. Although color baseline shifting eliminated aliasing in all subjects in this study, it is possible that double aliasing might occur when imaging is done at greater depth, at a higher transducer frequency, or in patients with very high cardiac output or a narrow LVOT. For an LVOT of diameter D imaged at depth d with ultrasound frequency f (and assuming a flat velocity profile through the narrowest portion), the critical instantaneous flow rate where uncorrectable aliasing will begin (Qmax) is given by πD2 c2/4df, where c is the speed of sound in blood. Expressing D and d in centimeters and f in megahertz, Qmax (in liters per second) becomes, in round numbers, 5D2/df (eg, 0.1 L/s for flow through a 1-cm LVOT imaged at a 20-cm depth with a 2.5-MHz transducer). Finally, although the algorithm is designed to be relatively tolerant of misalignment of the ultrasound beam with the flow stream, we did not systematically explore the validity of this assumption in the clinical setting. The accuracy of our clinical data and the in vitro results suggest, however, that misalignment is not a serious problem when imaging is done carefully. The in vitro study was limited by the known flat velocity profile of the EchoCal flow phantom; thus, the most innovative aspect of ACOM, its ability to deal with complex flow profiles, was not tested.
Clinical Implications and Future Directions
Our results suggest that this automated algorithm could be used to quantify cardiac output and stroke volume in most patients undergoing transthoracic echocardiographic evaluation, whether on an ambulatory or in-patient basis. Of these measurements, stroke volume is more important because cardiac output may be normalized by compensatory tachycardia, providing misleading reassurance to the clinician. In the intensive care setting, this has important cost and safety implications. Kaul et al25 demonstrated that simple 2D echocardiography provided most of the information from pulmonary artery catheterization. The additional hemodynamic information provided by this automated Doppler method will enhance its value. This is of particular relevance because a recent multicenter study has suggested that pulmonary artery catheterization increases hospital costs by 38% and actually worsens outcome.26
However, the principle of spatiotemporal velocity integration need not be confined to the LVOT. If flow through the mitral annulus could be similarly quantified, then combining this with aortic flow would permit easy quantification of mitral or aortic regurgitation as long as both were not present. Similarly, quantification of flow on the right side (tricuspid or pulmonary position) would permit quantification of shunt flow or as a forward reference standard to quantify mitral and aortic regurgitation when they coexist. Problems may exist with each of these approaches (such as the noncircularity of the mitral and tricuspid annulus), but they represent promising areas for future research. Similarly, it may be possible to apply this technology to transesophageal echocardiography, which would extend the utility of ACOM to the intraoperative arena. Unfortunately, in a large subset of patients, it may not be possible to visualize adequately the LVOT transgastrically to apply ACOM; a greater proportion of patients may have adequate images of the mitral annulus for inflow calculations, but such an approach will require extensive validation and perhaps modification of the ACOM software internal to the echo machine.
Automatic integration of numerical data within color Doppler flow fields is a feasible new method for quantifying cardiac output. It uses only one apical long-axis view, is simpler and faster, and requires fewer assumptions than alternative invasive and noninvasive approaches. This is a promising new approach to echocardiographic quantification that deserves further study and refinement.
Selected Abbreviations and Acronyms
|ACOM||=||automated cardiac output measurement|
|LVOT||=||left ventricular outflow tract|
|PWAO||=||pulsed Doppler measurement of aortic valve flow|
This study was supported in part by an equipment grant from Toshiba Corp.
This work was presented in part at the Sixth Annual Scientific Sessions of the American Society of Echocardiography, Toronto, Canada, June 14-16, 1995.
- Received September 12, 1996.
- Revision received October 14, 1996.
- Accepted October 16, 1996.
- Copyright © 1997 by American Heart Association
Branthwaite MA, Bradly RD. Measurement of cardiac output by thermal dilution in man. J Appl Physiol. 1986;24:434-438.
Fisher DC, Sahn DJ, Friedman MJ, Larson D, Valdez-Cruz LM, Horowitz S, Goldberg SJ, Allen HD. The effect of variations on pulsed Doppler sampling site on calculation of cardiac output: an experimental study in open-chest dogs. Circulation. 1983;67:370-376.
Lewis JF, Kuo LC, Nelson JG, Limmacher MC, Quinones MA. Pulsed Doppler echocardiographic determination of stroke volume and cardiac output: clinical validation of two new methods using the apical window. Circulation. 1984;70:425-431.
Sarano ME, Bailey KR, Seward JB, Tajik AJ, Krohn MJ, Mays JM. Quantitative Doppler assessment of valvular regurgitation. Circulation. 1993;87:841-848.
Sun JP, Stewart WJ, Yang XS, Lee KS, Sheldon WS, Thomas JD. Automated echocardiographic quantification of left ventricular volumes and ejection fraction: validation in the intensive care setting. J Am Soc Echocardiogr. 1995:8:29-36.
Huntsman LL, Stewart DK, Barnes SR, Franklin SB, Colocousis JS, Hessel EA. Noninvasive Doppler determination of cardiac output in man: clinical validation. Circulation. 1983;67:593-602.
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, for the SUPPORT Investigators. The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA. 1996;276:889-897.