Pulmonary regurgitation (PR) is a common postoperative sequel in patients with tetralogy of Fallot (TOF). Noninvasive quantification of the PR may be of major clinical importance in the timing of reoperation since PR-induced right ventricular (RV) volume overload impairs exercise ability in patients after repair of TOF and may predispose a patient to the development of RV dysfunction (1- 3).

The noninvasive techniques that are presently used in quantifying PR include a pulsed Doppler method (1,4) and a velocity encoded magnetic resonance imaging (5). However, the quantitative pulsed Doppler method requires meticulous care in measuring the area of flow tract, and the pulsed Doppler method assumes a flat velocity profile during forward and reverse flow as well as a constant flow area, assumptions which are rarely correct (6- 9). The high cost, long acquisition time and nonportability of magnetic resonance imaging instrumentation limit the use of this technique in clinical practice.

Recently, a technique has been developed for automated cardiac flow measurement (ACM) based on spatiotemporal integration of digital color Doppler (DCD) velocities, which provides accurate estimates of cardiac output in in vitro studies (10- 13) and in clinical setting (14,15). We hypothesized that: 1) the DCD method could estimate the amount of PR and 2) the DCD cardiac output measurement method using average values from two orthogonal planes could improve the accuracy of quantifying PR. To test this hypothesis, we performed the DCD studies for forward and reversed flow using two orthogonal planes for determining pulmonary regurgitant volumes (RSVs) and regurgitant fractions (RFs) in an in vitro model mimicking postoperative TOF without peripheral pulmonary stenosis and in a chronic animal model.

Our repaired TOF model consisted of a rubber tube and a more compliant latex tube 18-mm in diameter. The rubber tube and the compliant latex tube were modeled to simulate the RVOT and main pulmonary artery (MPA), respectively. At the outlet of the rubber tube, a restricting ring with a 10-mm inner diameter was inserted so as to produce mild stenosis mimicking a pulmonary annulus with no valve function. An external constrictor also was placed around the end of the rubber tube to alter its shape from circular to a more clinically relevant ovoid geometry. The diameter of major axis and minor axis of ovoid is about 21 mm and 15 mm, respectively (Figure 1). A solution of water mixed with 1% cornstarch was used as fluid medium to improve echocardiography reflectivity. The repaired TOF model was connected to a computer controlled programmable pulsatile pump (UHDC flow system, Model PFS A-3-1-G-1, London, Ontario, Canada). A fixed pump rate of 50 beats/min was used. Five different sinusoidal waveforms were used to obtain different RFs. For each waveform, different forward and reversed stroke volumes were obtained by changing peak flow rate and the waveform program for the pump. Instantaneous forward and reversed flow were measured by an ultrasonic probe (Model 16 NB272, Transonic system Inc, Ithaca, New York) and T201-2-channel ultrasonic blood flow meter (Model T 106X, Transonic System Inc., Ithaca, New York) and were recorded at 25 mm/s on chart paper. The forward stroke volume (FSV) per beat and RSV per beat were obtained by planimetry of the flow signal recordings on the strip chart paper to yield as reference values. Regurgitant fraction was calculated as the ratio of the RSV per beat to the FSV per beat. Three consecutive cardiac cycles were measured and averaged. Thirteen different FSVs and RSVs were examined.

Three to six months before the hemodynamic and ultrasonic study sessions, PR was surgically created in six juvenile sheep weighing 18 to 57 kg. All operative and animal management procedures were approved by the Animal Care and Use Committee of the National Heart, Lung, and Blood Institute.

To obtain pulmonary RSV and RF, two electromagnetic flow (EM) probes (model EP455, Carolina Medical Electronics, Inc., King, North Carolina) were placed, one around the pulmonary artery just above the pulmonary valve and the second around the skeletonized ascending aorta distal to the coronary ostia and proximal to the brachiocephalic trunk. Both flow probes were connected to flow meters (model FM501, Carolina Medical Electronics) and interfaced to the physiological recorder (Gould ES 2000, Cleveland, Ohio). All hemodynamic data were recorded at a paper speeds of 250 mm/s. Four consecutive cardiac cycles were analyzed for each hemodynamic determination as describe elsewhere (16,17).

For determination of pulmonary and aortic flow volumes, the integrals of instantaneous pulmonary and aortic flows over time were determined by planimetry of the flow signal recordings. The problem of the zero baseline drift was managed as previously described (18) so that the baseline value for the pulmonary flow recordings was adjusted until the forward minus the reversed pulmonary flow volumes equaled the aortic forward flow volumes. The RF was calculated as diastolic, reverse pulmonary flow volume per minute divided by forward pulmonary flow volume per minute.

A total of 24 stable hemodynamic states were obtained in the six sheep by altering preload or afterload using blood transfusion, sodium nitroprusside and angiotensin II (Peptide Institute Inc., provided by Tanabe Seiyaku Co., Osaka, Japan).

Color Doppler imaging was performed with a Toshiba Power Vision SSA-380A (Toshiba Corp., Tokyo, Japan) equipped with software for ACM. The principle of the ACM as a DCD method is described in detail elsewhere (10,12- 14,15). A 5MHz omniplane transesophageal probe was placed in a water bath adjacent to the RVOT portion of our model to image parallel to flow. Color Doppler image loops were obtained in the two orthogonal planes corresponding to major and minor axes of the RVOT lumen (Figures 1, 2). The transducer was located and fixed about 2.5 cm away from the “pulmonary annulus,” and the central Doppler beam was oriented parallel to flow.

In the animal study, the color Doppler imaging was performed using a 5 MHZ sector probe placed directly on the RV, half way to the apex and angled superiorly. An oblique outflow view corresponding to the clinically conventional subcostal anteriorly angled outflow view and an orthogonal view corresponding to the conventional parasternal long axis view were used to image the RVOT and MPA (Figure 3). For both the in vitro and the in vivo study, an optimal gain setting was set so as to avoid random color noise in the nonflow areas. Wall filtering was minimized so as not to lose low velocity information. A pulse repetition frequency 4.5 to 6.0 KHz for the in vitro study and 6 to 8.0 KHz for the animal study was used for color Doppler scanning. To avoid aliasing, baseline shift was performed where necessary so that unaliased flow was imaged in the area for DCD computation throughout systole or diastole. A narrow color sector was chosen so as to maximize frame rate, and frame rate ranged from 18 to 28 frames/s. Digital color Doppler measurements were obtained on board the Toshiba system using their ACM software.

For determination of the FSVs, the region of interest was placed below the first aliasing boundary of the flow convergence region in the rubber tube mimicking RVOT or above the pulmonary valve in the RVOT in the animals. On a selected beat, a total of 7 to 17 frames during systole were selected from the cine loop manually. For determination of the RSVs on the same loop, the region of interest was positioned in the distal compliant latex tube or the MPA in the animals, and a total of 3 to 13 frames during diastole were selected (Figures 2, 3). Three or four determinations of both FSVs and RSVs of separate beats were averaged for each flow state. The RF was calculated as the ratio of the RSV to the FSV.

To evaluate the effect of interobserver variability on measurement of FSV and RSV, two investigators independently performed measurements on 10 randomly selected flow conditions in our in vitro model. The two independent investigators were blinded to the flow rates of the images selected or the reference flow meter data.

Data are expressed as mean ± standard deviation. Correlation between the DCD computed data and reference data were assessed by simple and multiple linear regression analyses (19). Agreement between the DCD computed data and reference data was tested according to the Bland-Altman method (20). The DCD computed data were compared with the references using paired Student t test. For comparison of mean difference (DCD data-reference data) between three different planes, analysis of variance with Scheffé F test was used. A value of p < 0.05 was considered statistically significant.

Peak pressure gradients across the pulmonary annulus determined by continuous wave Doppler ranged from 10 to 25 mm Hg (average, 14 ± 6 mm Hg), and the pulmonary RFs obtained by flow meter covered the ranges of mild to severe regurgitation from 10% to 71% (average, 41 ± 19%).

Correlation and mean difference between the DCD method in two orthogonal planes and flow meter method for measuring FSVs, RSVs and RFs are shown in (Table le1) and (Figure 4). The simple regression analysis showed that the FSVs and RSVs derived by the DCD method in either plane and those using an average of two orthogonal planes correlated well with the references (Table le1). In the horizontal plane corresponding to the major axis, the DCD method overestimated the actual FSVs (mean degree of overestimation = 139 ± 12%, p < 0.0001) and RSVs (mean degree of overestimation = 129 ± 20%, p = 0.0005), but in the paired orthogonal plane corresponding to the minor axis, this method underestimated the actual FSVs (mean degree of underestimation = 68 ± 11%, p < 0.0001) and RSVs (mean degree of underestimation = 68 ± 14%, p = 0.0007). However, there was no significant difference in the FSVs and RSVs between the DCD computed data using the average of two orthogonal planes and reference data (p = 0.16 for the FSV and p = 0.84 for the RSV). As a result, these measurements of the pulmonary RFs by the DCD method agreed with the references using all implementation methods (the major, the minor axis and the average of two orthogonal planes) ((Figure 4), Table le1). None of the RFs obtained by three different implementations were significantly different from the references (p = 0.15 for the major axis, p = 0.72 for the minor axis and p = 0.84 for the average of two orthogonal planes), and the mean difference of RFs was also not different between three different planes (p = 0.43).

Thus, the FSVs and RSVs were more accurately estimated by the DCD method using the average of two orthogonal planes than by using any single plane, although either plane appears to be adequate for the measurement of RFs.

Peak velocities across the pulmonary annulus determined by continuous wave Doppler ranged from 0.6 to 3.1 m/s (average, 1.4 ± 0.6 m/s). One sheep had mild, but significant, pulmonary stenosis in which the peak pressure gradients across the pulmonary annulus determined by continuous wave Doppler ranged from 16 to 36 mm Hg (average, 31 ± 6 mm Hg). Pulmonary RFs obtained by the EM method were within clinically relevant ranges from 9.8% to 75% (average, 37 ± 18%). Heart rate ranged from 88 to 136 beats/min (average, 104 ± 12 beats/min).

Correlation and mean difference between the DCD method in two orthogonal planes and the EM method for measuring FSVs, RSVs and RFs are shown in (Table le2) and (Figure 5). Multiple regression analysis showed that the FSVs and RSVs derived by the DCD method in a single plane correlated with those obtained by the EM method (Table le2). In the oblique outflow view, the DCD method overestimated the actual FSVs (mean degree of overestimation = 148 ± 21%, p < 0.0001), but in the long axis view corresponding to the orthogonal plane, this method underestimated the actual FSVs (mean degree of underestimation = 74 ± 12%, p < 0.0001). The DCD method significantly overestimated the actual RSVs in either plane (mean degree of overestimation = 136 ± 64% for the oblique outflow view, p = 0.006 and mean degree of overestimation = 132 ± 66% for the long axis view, p = 0.04). As a result of these measurements, the RFs obtained by the DCD method using either single plane also correlated with reference data but with wide variability, especially in the long axis view (Figure 5). In the oblique outflow view, the DCD method underestimated actual RFs (mean degree of underestimation = 90 ± 35%, p = 0.0009), but in the long axis view, it overestimated actual RFs (mean degree of overestimation = 191 ± 126%, p < 0.0001) (Table le2).

When the average of the two orthogonal planes was used, the FSVs and RSVs derived by the DCD method correlated and agreed more closely with the reference data compared with those derived by the DCD method using the single plane. However, the DCD method using the average of two orthogonal planes significantly overestimated the actual FSVs and RSVs (mean degree of overestimation = 111 ± 10% for the FSV, p < 0.0001 and mean degree of overestimation = 134 ± 57% for the RSV, p < 0.0001) (Table le2). As a result of these measurements, there was excellent correlation and agreement between the RFs by the DCD method using the average of two orthogonal planes and by the EM method ((Figure 5), Table le2). There was no significant difference in the RFs between the DCD computed data using the average of two orthogonal planes and reference data (p = 0.06). Thus, the FSVs, RSVs and RFs were more accurately estimated by the DCD method using the average of two orthogonal planes than by using any single plane.

There was good agreement between the two independent observers’ determinations of RSVs (r = 0.98, p < 0.0001, mean absolute difference = 0.06 ± 0.71 ml/beat) and RFs (r = 0.93, p < 0.0001, mean absolute difference = 1.11 ± 5.59%) for the in vitro data.

The DCD method we used has several advantages over the pulsed Doppler method for quantifying PR. First, since the area of flow is computed directly, it does not require a separate step and separate views to measure the size of vessel or the area of the flow tract, which is a major source of error in the pulsed Doppler method (6- 7). The DCD method accounts for the dynamically changing vessel size and flow profile through frame by frame temporal integration during the systolic phase or diastolic phase of flow. We demonstrated that RSVs and RFs could be accurately estimated even in the presence of mild pulmonary stenosis by this method, although it was necessary to use two orthogonal planes in the case of noncircular geometry, such as in our model. Another advantage of this DCD method is that it is easy and quick to implement.

Although the DCD method does not assume either a flat flow profile or constant flow area, it does assume a circular shaped flow area and an axial symmetry of the spatial flow profile (10,12- 14,15). Noncircular geometry may be present in the RVOT or MPA in postoperative patients with TOF as in our model. Our study demonstrated that the DCD method using the average of two orthogonal planes could mitigate the limitations of assuming axial symmetry of flow area and velocity profile, providing accurate estimation of pulmonary RSVs and RFs. In our in vitro model, the DCD method using a single plane overestimated the actual FSVs and RSVs, and the paired orthogonal plane underestimated them. As a result, the DCD method that used the average of two orthogonal planes estimated more accurately the actual FSVs and RSVs than it did by using a single plane. In our animal model, the DCD method using a single plane corresponding to the oblique outflow view overestimated FSVs, and the paired orthogonal plane corresponding to the long axis view underestimated them. Both single planes (oblique outflow view and long axis view) methods overestimated the actual RSVs. As a result, the DCD method using the average of two orthogonal planes yielded a better estimate of the actual FSVs, RSV and RFs than it did using the single plane, although it slightly overestimated the actual FSVs and RSVs. In our in vitro model, the shape of the cross-sectional flow area of the MPA was ovoid, and the spatial velocity distribution of the regurgitant flow in the ovoid flow area was almost uniform with the regurgitant orifice located in the center of the tube mimicking the pulmonary annulus. In fact, although the diameters were different, the shape of velocity profiles of regurgitant flow was almost similar in two orthogonal planes in our in vitro model (Figure 2). In contrast, in our animal model, the cross-sectional flow area of the MPA was different from that in the in vitro model: near circular as opposed to ovoid. The spatial velocity distribution of the regurgitant flow, however, was different in the in vivo model both as a result of variable orifice position and MPA curvature. The diameters of the velocity profiles of the regurgitant flow were similar in the oblique outflow view and the long axis view at the same phase of diastole, but the shapes of the velocity profiles were different (Figure 3). In addition, although the shapes of the cross-sectional flow area of the RVOT and MPA were the same in our vitro model, they differed substantially in our animal model (Figure 3) and would probably differ substantially in patients. In the in vivo study, the reason for overestimation in RSVs obtained by either single plane, and in RSVs and FSVs obtained by the average of two orthogonal planes, may be related to the difficulty in obtaining exact orthogonal planes and, in particular, in not identifying the actual minor axis. Recently, in preliminary studies, we have demonstrated that the three-dimensional DCD method without geometric assumption so as to calculate the flow volumes should be more objective and less susceptible to error (21).

Some patients with repaired TOF may have peripheral pulmonary stenosis, not simulated in this study. In vitro and clinical studies have shown the effect of color gain settings, wall filters, frame rates and pulse repetition frequency on any color Doppler method (10,12- 14,15). In clinical settings, it may not always be possible to have the high quality images we had, and it may be difficult to image parallel to flow as we did in our in vitro model. The DCD algorithm as implemented is relatively tolerant of mild degrees of misalignment of the Doppler beam orientation with flow direction (14). Fortunately, the angle between forward and regurgitant flow direction in the RVOT and Doppler beam orientation can be minimized by appropriate transducer position from the oblique outflow view (subcostal or parasternal) or long axis view of RVOT in patients.

Our study has demonstrated in our models of postoperative TOF that the DCD method is potentially clinically useful for quantifying pulmonary RSVs and RFs.

The authors thank Judy Schultheis for her assistance with the preparation of the manuscript.