Background
The objective of this study was to test the feasibility, accuracy, and reproducibility of the assessment of right ventricular (RV) volumes and ejection fraction (EF) using real-time three-dimensional echocardiographic (RT3DE) imaging in patients with congenital heart disease (CHD), using cardiac magnetic resonance (CMR) as a reference.
Methods
RT3DE data sets and short-axis cine CMR images were obtained in 62 consecutive patients (mean age, 26.9 ± 10.4 years; 65% men) with various CHDs. RV volumetric quantification was done using semiautomated 3-dimensional border detection for RT3DE images and manual tracing of contours in multiple slices for CMR images.
Results
Adequate RV RT3DE data sets could be analyzed in 50 of 62 patients (81%). The time needed for RV acquisition and analysis was less for RT3DE imaging than for CMR ( P < .001). Compared with CMR, RT3DE imaging underestimated RV end-diastolic and end-systolic volumes and EF by 34 ± 65 mL, 11 ± 55 mL, and 4 ± 13% ( P < .05) with 95% limits of agreement of ±131 mL, ±109 mL, and ±27%, as shown by Bland-Altman analyses, with highly significant correlations ( r = 0.93, r = 0.91, and r = 0.74, respectively, P < .001). Interobserver variability was 1 ± 15%, 6 ± 17%, and 8 ± 13% for end-diastolic and end-systolic volumes and EF, respectively.
Conclusion
In the majority of unselected patients with complex CHD, RT3DE imaging provides a fast and reproducible assessment of RV volumes and EF with fair to good accuracy compared with CMR reference data when using current commercially available hardware and software. Further studies are warranted to confirm our data in similar and other patient populations to establish its use in clinical practice.
Because of improved surgical techniques and medical care, a growing number of patients with congenital heart disease (CHD) survive into adulthood. Right ventricular (RV) dysfunction is a common problem in these patients, associated with significant morbidity and mortality. Therefore, regular assessments of RV function in these patients are essential for clinical management. Accurate and accessible tools are needed to monitor RV function, which will lead to better timing of surgical reintervention and medical therapy, ultimately with better survival.
Currently, cardiac magnetic resonance (CMR) imaging is the standard for the quantification of RV volumes and ejection fraction (EF). However, CMR is not widely available and is expensive, acquisition and offline analysis are time-consuming, and a substantial number of patients with CHD have pacemakers or implantable cardioverter-defibrillators and thus (relative) contraindications for CMR.
In routine clinical practice, two-dimensional (2D) echocardiography is most commonly used for the noninvasive evaluation of cardiac function. However, it is well known that because of the complex cardiac geometry, 2D echocardiography has important limitations in the assessment of left ventricular and in particular RV volumes and EF. To overcome the problem of geometric assumptions and apical foreshortening, real-time three-dimensional echocardiographic (RT3DE) imaging was developed, which allows the fast acquisition of a pyramidal data set that contains the entire right ventricle. In experimental settings, this imaging modality was successfully applied for RV volume and EF calculation in both healthy volunteers and pediatric patients with CHD. These studies were limited by selected, nonconsecutive subjects or the use of older or not commercially available data analysis methods. In vitro validation has been done to investigate variables influencing the accuracy of RV RT3DE imaging, in which optimal gain settings and long-axis tracings were found to significantly affect RV volumes.
The purpose of our study was to determine whether current commercially available hardware and software for the assessment of RV volumes and EF with RT3DE imaging can be applied in routine clinical CHD practice.
Methods
Study Population
RT3DE full-volume acquisition of the right ventricle was performed on 62 consecutive patients with complex and/or surgically repaired CHD. They were in sinus rhythm and represented a wide range of RV EFs. The patients were referred for CMR for the quantitative analysis of cardiac function for clinical reasons and underwent RT3DE examinations within 2 hours of CMR to pursue comparable loading conditions. The medical ethics committee approved this study, and written informed consent was obtained from all patients and/or their parents (if required).
RT3DE Protocol and Data Analysis
Data Acquisition
RT3DE harmonic imaging was performed using the iE33 ultrasound system (Philips Medical Systems, Best, The Netherlands) equipped with an X3-1 matrix array transducer, with the patient in the left lateral decubitus position. To encompass the entire right ventricle into the RT3DE data set, a full-volume scan was acquired from a modified apical transducer position in harmonic mode from 7 R wave–gated subvolumes during a single end-expiratory breath hold. The output therefore was not truly real time but reconstructed from 7 subvolumes. The depth and angle of the ultrasound sector were adjusted to a minimal level still encompassing the entire right ventricle. Before each acquisition, images were optimized for endocardial border visualization by modifying the time gain and compression and increasing the overall gain. An average of 3 data sets was acquired per patient, to ensure optimal data sets without motion artifacts that may have occurred during the acquisition. The mean volume rate was 25 frames per cardiac cycle (range, 14-36). The data sets were digitally exported to a TomTec server (TomTec Imaging Systems, Unterschleissheim, Germany) connected to a terminal workstation for further analyses.
Data Analysis
The digital RT3DE RV data sets were analyzed offline using four-dimensional RV-Function version 4.0 (TomTec Imaging Systems) by an investigator blinded to the results of the CMR measurements (H.B.Z.). This software performs 3D semiautomated border detection of RV volumes over 1 cardiac cycle. It uses a physics-based modeling algorithm that makes no assumptions regarding RV geometry. Analysis of a RT3DE data set was judged possible when both the apex and the lateral wall were visible in the 4-chamber view, allowing adequate tracing of the endocardial border. Analysis of the data set was considered good when the RV anterior wall or the outflow tract was visible.
Once the TomTec analysis program starts, the screen displays a short-axis view (top), an apical 4-chamber view (left), and a coronal view (right) ( Figure 1 ). RV quantification starts by definition of the correct apical 4-chamber view, with avoidance of RV foreshortening. Then it is made sure that the displayed vertical lines are in the middle of the tricuspid valve and apex in both the apical 4-chamber and the coronal views. Subsequently, the horizontal line displayed on the apical 4-chamber view is moved to the level of the atrioventricular valves, resulting in a view of both valve orifices at the short-axis view. The next step is to place landmarks in both the tricuspid and mitral valve orifices. Then the horizontal line is moved up to the left ventricular apex, resulting in a short-axis view of the apex at the top. A third landmark is placed at the apex. Afterward, the end-diastolic (largest RV volume) and end-systolic (smallest RV volume) phases are identified. Then the endocardial border contours are drawn onto the apical 4-chamber view in both the end-diastolic and end-systolic frames. At the side of these frames, a movie of the current view is displayed to facilitate detection of the endocardial contours that are drawn in the still frames. The contours are adjusted as close as possible to the endocardial border. Trabeculae are mostly excluded from the volume, because of poor differentiation of trabeculae and myocardium ( Figure 2 ). On the basis of these contours traced from the apical 4-chamber view, 2 reference markers are extrapolated onto the sagittal view. Onto this sagittal view, endocardial border contours are traced, with care taken to include the 2 extrapolated reference markers in the end-diastolic and end-systolic frames. Hereafter, contours are drawn in the coronal view, again with attention to include the 3 reference markers that were extrapolated from the 4-chamber and sagittal views. Hereafter the software automatically delineates the RV endocardial border from the end-diastolic and end-systolic phases, and by sequential analysis, the software creates an RV mathematic dynamic 3D model, referred to as a “beutel,” that represents changes in the RV cavity over the cardiac cycle. From this beutel, RV end-diastolic volume (EDV), end-systolic volume (ESV) and EF (defined as [EDV − ESV]/EDV × 100) are derived. The background computation of the RV volumetric data is described in detail by Iriart et al. Hereafter, manual correction of the contours can be done in any cross-section or phase of the cardiac cycle ( Figure 3 , Video 1 ; 
view video clip online.).
RT3DE Reproducibility Analysis
RT3DE measurements were repeated in 25 randomly selected patients by the same observer after ≥1 month from the original measurements to obtain intraobserver values blinded to CMR results. A second observer (J.S.M.) repeated the measurements in 25 randomly selected patients for interobserver comparison.
CMR
CMR images were acquired using a Signa 1.5-T scanner (GE Medical Systems, Milwaukee, WI). Patients were positioned in the supine position with dedicated phased-array cardiac surface coils placed over the thorax. The CMR protocol included cine steady-state free precision sequences in short-axis planes to assess the ventricular volumes. Electrocardiographic gating and repeated breath holds were applied to minimize the influence of cardiac and respiratory motion.
Ventricular volumes were measured from a multisection image set of 8 to 12 contiguous slices parallel to the plane of the atrioventricular valves covering the full length of both ventricles. Imaging parameters were as follows: slice thickness, 7 to 10 mm; slice gap, 0 mm; field of view, 280 to 370 mm; phase field of view, 0.75; matrix size, 160 × 128 mm; repetition time, 3.5 ms; echo time, 1.5 ms; and flip angle, 45°.
Three months after analysis of the RT3DE data sets, one physician (H.B.Z.) analyzed the CMR ventricular volumetric data set quantitatively on a commercially available Advanced Windows workstation (GE Medical Systems) using Advanced Windows version 5.1 of the MR Analytical Software System (Medis Medical Imaging Systems, Leiden, The Netherlands). All CMR data sets were analyzed in a blinded way to prevent influence of the analyzer by the results of RT3DE imaging. Using manual detection of endocardial borders in end-systole and end-diastole, the RV EDV, ESV, and EF were calculated with exclusion of trabeculae, as described by Alfakih et al. In a random subgroup of 15 patients, we determined the volume of the trabeculae on the CMR images by subtraction of the volume with inclusion of trabeculae and the volume with exclusion of trabeculae.
Statistical Analysis
Statistical analysis was performed using SPSS version 15.0 (SPSS, Inc, Chicago, IL). Categorical variables are summarized as numbers and percentages. Continuous variables are presented as mean ± SD, and differences were analyzed using Student’s t tests. Linear regression analysis with Pearson’s correlation coefficient was used to evaluate the relation between RT3DE imaging and CMR. The agreement between RT3DE and CMR measurements was evaluated using Bland-Altman analysis by calculating the bias (mean difference) and the 95% limits of agreement (2 SDs around the mean difference). Paired t tests were used to analyze the significance of biases in volumes and EF between RT3DE imaging and CMR. The reproducibility of RT3DE imaging was evaluated by calculating the intraobserver and interobserver variability by a variation coefficient, which was defined as the absolute difference expressed as a percentage of the mean of both values. All statistical tests were two sided, and a P value < .05 was considered statistically significant.
Results
Population Characteristics
Of the 62 consecutively enrolled patients (mean age, 26.9 ± 10.4 years; 65% men), 12 (19%) had to be excluded because of insufficient image quality. Age, sex, weight, and type of CHD in these patients were not different from these variables in patients with sufficient image quality. The final cohort comprised 50 patients with a wide range of RV morphology and loading conditions. These included normal right ventricles in left-sided heart disease such as aortic valve pathology, subpulmonary right ventricles facing pressure overload as in pulmonary valve stenosis, volume overload as in atrial septal defects, and right ventricles with severely dilated outflow tracts as in tetralogy of Fallot. Patient characteristics are presented in Table 1 .
| Variable | Value |
|---|---|
| Clinical data | |
| Age (y) | 26.3 ± 10.5 |
| Men | 68% |
| Heart rate (beats/min) | 72 ± 13 |
| Body mass index (kg/m 2 ) | 22.9 ± 3.9 |
| Body surface area (m 2 ) | 1.8 ± 0.3 |
| Pathology | |
| Tetralogy of Fallot | 21 |
| Pulmonary stenosis ± ventricular septal defect | 5 |
| Pulmonary atresia ± ventricular septal defect | 3 |
| Transposition of the great arteries, atrial switch | 4 |
| Anomalous pulmonary venous drainage | 1 |
| Tricuspid insufficiency | 1 |
| Atrial septal defect | 1 |
| Aortic valve pathology | 10 |
| Transposition of the great arteries, arterial switch | 3 |
| Cardiomyopathy | 1 |
Volume Analysis
Table 2 shows the mean RV EDV, ESV, and EF for RT3DE imaging and CMR. Linear regression analysis ( Table 2 , Figure 4 ) showed acceptable correlations between RT3DE imaging and CMR for EDV ( r = 0.93, y = 0.76 x + 19 mL, P < .001), ESV ( r = 0.91, y = 0.71 x + 22 mL, P < .001), and EF ( r = 0.74, y = 0.60 x + 16%, P < .001). Bland-Altman analysis showed mean differences of 34 mL for EDV, 11 mL for ESV, and 4% for EF and 95% limits of agreement of ±65 mL for EDV, ±55 mL for ESV, and ±13% for EF (all P values < .05; Figure 5 ). The mean volume of the trabeculae on CMR images was 19 ± 13 mL in systole and diastole.
