The aim of this study was to validate a novel real-time three-dimensional echocardiographic (RT3DE) analysis tool for the determination of right ventricular volumes and function in unselected adult patients.
A total of 100 consecutive adult patients with normal or pathologic right ventricles were enrolled in the study. A dynamic polyhedron model of the right ventricle was generated using dedicated RT3DE software. Volumes and ejection fractions were determined and compared with results obtained on magnetic resonance imaging (MRI) in 88 patients with adequate acquisitions.
End-diastolic, end-systolic, and stroke volumes were slightly lower on RT3DE imaging than on MRI (124.0 ± 34.4 vs 134.2 ± 39.2 mL, P < .001; 65.2 ± 23.5 vs 69.7 ± 25.5 mL, P = .02; and 58.8 ± 18.4 vs 64.5 ± 24.1 mL, P < .01, respectively), while no significant difference was observed for ejection fraction (47.8 ± 8.5% vs 48.2 ± 10.8%, P = .57). Correlation coefficients on Bland-Altman analysis were r = 0.84 (mean difference, 10.2 mL; 95% confidence interval [CI], −31.3 to 51.7 mL) for end-diastolic volume, r = 0.83 (mean difference, 4.5 mL; 95% CI, −23.8 to 32.9 mL) for end-systolic volume, r = 0.77 (mean difference, 5.7 mL; 95% CI, −24.6 to 36.0 mL) for stroke volume, and r = 0.72 (mean difference, 0.4%; 95% CI, −14.2% to 15.1%) for ejection fraction.
Right ventricular volumes and ejection fractions as assessed using RT3DE imaging compare well with MRI measurements. RT3DE imaging may become a time-saving and cost-saving alternative to MRI for the quantitative assessment of right ventricular size and function.
Experimental studies on open-chest canine models in the 1940s and 1950s suggested that even extensive damage to the right ventricular (RV) wall does not significantly influence circulation. In fact, the right ventricle is not a mere bystander but plays an important role in cardiac hemodynamics. In the 1980s and 1990s, it was demonstrated that RV dysfunction is an independent prognostic factor in a variety of diseases. However, because of the complex shape of the right ventricle, the reliable quantitative assessment of its size and function has remained an unresolved issue. Conventional echocardiographic methods do not enable RV volume measurements and provide only surrogates of ejection fraction. Hence, methods that allow the chamber to be portrayed in three dimensions appear likely to be the future techniques of choice. Various three-dimensional (3D) echocardiographic techniques have been applied for the evaluation of the right ventricle. With the advent of real-time 3D echocardiographic (RT3DE) imaging, a technique is available that enables the rapid acquisition of 3D data sets. Initial studies using RT3DE imaging were based on the provisional use of software dedicated to the left ventricle. Recently, an analysis tool especially designed for the reconstruction of a dynamic polyhedron RV model has been developed. Two initial validation studies in groups of predominantly pediatric patients yielded good agreement with magnetic resonance imaging (MRI). The aim of the present study was to assess for the first time the accuracy and reproducibility of the novel technique in a large cohort of unselected adult patients.
One hundred adult patients who underwent clinically indicated cardiac MRI were included prospectively from October 2007 to December 2008. The patients were included solely on the basis of indications for MRI, so the study population was unbiased regarding echocardiographic image quality. In 97% of the patients, RT3DE imaging and MRI were performed on the same day to minimize the impact of changing loading conditions. In the remaining 3 patients, RT3DE acquisitions were performed <24 hours after MRI. Informed consent was obtained from all patients. The study protocol was approved by the local institutional review committee.
RT3DE Data Set Acquisition
All 3D data sets were acquired using a Philips iE33 ultrasound system equipped with a matrix-array X3-1 transducer (Philips Medical Systems, Andover, MA). Acquisitions were performed using a modified apical view to enable full coverage of the entire right ventricle by the pyramidal volume, with particular attention to the upper anterior wall and RV outflow tract ( Figure 1 , Video 1 ; view video clip online). The optimal position of the probe was controlled by stepwise 360° rotation of the modified apical view. Sector size and depth were chosen carefully to achieve the highest possible frame rate. Then a data set was recorded over 7 cardiac cycles. An average of 4 to 6 data sets were acquired per patient, and the data set with the highest image quality was used for further analysis.
RT3DE Data Analysis
Three-dimensional data sets were analyzed offline using the novel dedicated software (4D RV-Function CAP 1.1; TomTec Imaging Systems, Inc, Unterschleissheim, Germany) and a software platform for data management (Research Arena 2.0; TomTec Imaging Systems, Inc). In brief, the work flow of the RV analysis software is as follows: Step 1, view adjustment: within the 3D data set, 3 orthogonal main cut planes are selected ( Figure 2 , Video 2 ; view video clip online), and the observer can define the end-diastolic and end-systolic frames within the sequence as well as several landmarks. Step 2, setting of initial contours: on the basis of the initial view adjustment and the landmarks, the program automatically provides 4-chamber, sagittal, and coronal views of the right ventricle. The observer draws the end-diastolic and end-systolic contours manually in each view. Contouring the endocardium of the RV outflow tract is the most challenging task during offline data analysis. In case of suboptimal endocardial visualization of this region, contours of the RV outflow tract must be extrapolated on the basis of the adjacent endocardium with improved visualization to close the envelope and to advance within the predefined steps of the software. Trabeculations were included in the RV volume. Step 3, contour revision: the application defines a dynamic polyhedron model of the right ventricle on the basis of the initial contours. This model is automatically adapted to the endocardial surface of the ventricle over all frames of the cardiac cycle. If necessary, the observer can manually correct the contours ( Figure 3 , Video 3 ; view video clip online). Step 4, RV analysis: finally, the RV analysis display offers the dynamic model and a table with the values of RV volumes and function ( Figure 3 , Videos 3-5 ; view video clips online).
The MRI examinations were performed using a 1.5-T magnet (Magnetom Avanto or Espree; Siemens Medical Solutions, Forchheim, Germany) equipped with a phased-array body coil in supine position at end-inspiration. The heart was imaged with a prospectively gated, segmented (14 segments), steady-state free precession cine sequence (acceleration factor, 2; repetition time, 2.7 ms; echo time, 1.219; flip angle, 51°; matrix size, 156 × 192; bandwidth, 930 Hz/pixel). Spatial resolution was 1.9 × 1.9 in plane, with a slice thickness of 6 mm, and temporal resolution was 16 frames/cardiac cycle. Images were acquired in successive short-axis and 4-chamber views. Images were then processed with dedicated software (Argus; Siemens Medical Solutions) by contouring the endocardium in end-diastole and end-systole on every short-axis slice. The slices to be included in the calculations were identified by reviewing the images in cine mode to control for motion of the inflow and outflow tracts using standard analytic definitions. On the basis of these contours, RV stroke volume and ejection fraction were calculated by the summation-of-discs method ( Figure 4 , Video 6 ; view video clip online).
MRI examinations were performed by a team of radiologists and cardiologists at our hospital and measured by one observer solely. The selected RT3DE data set was evaluated by two different observers. For intraobserver variability, one of the observers performed a second RT3DE evaluation ≥30 days after the first evaluation. All measurements were done in a blinded fashion.
Data are presented as mean ± SD for continuous variables and as absolute numbers and relative percentages for categorical variables. Paired data were compared using Wilcoxon’s signed-rank test. Relations between two methods, two measurements, or two observers were determined using linear regression analysis, and the respective intraclass correlation coefficient was calculated. Intraobserver and interobserver agreement and agreement between methods were assessed using Bland-Altman analysis. Significance was defined as a two-sided P value < .05. SPSS version 16.0 (SPSS, Inc, Chicago, IL) was used for statistical analysis.
A total of 100 patients (mean age, 49 ± 16 years) were enrolled. Baseline characteristics of the study population are given in Table 1 . Eight RT3DE data sets (8%) and 3 MRI data sets (3%) were excluded because of insufficient image quality, and 1 MRI data set (1%) could not be analyzed because of technical failure. Thus, RT3DE and MRI data were compared in 88 patients (88%).
|Age (y)||50 ± 16 (18-79)|
|Height (cm)||175 ± 8 (154-195)|
|Weight (kg)||79 ± 14 (43-115)|
|Body mass index (kg/m 2 )||26 ± 4 (17-35)|
|Systolic blood pressure (mm Hg)||126 ± 15 (100-205)|
|Diastolic blood pressure (mm Hg)||77 ± 11 (51-97)|
|Heart rate (beats/min)||68 ± 14 (41-122)|
|Number of frames (RT3DE imaging)||23 ± 5 (11-38)|
|Ischemic heart disease||19%|
|Arrhythmogenic right ventricular cardiomyopathy||13%|
RT3DE Imaging Versus MRI
End-diastolic, end-systolic, and stroke volumes were slightly lower on RT3DE imaging than on MRI (124.0 ± 34.4 vs 134.2 ± 39.2 mL, P < .001; 65.2 ± 23.5 vs 69.7 ± 25.5 mL, P = .02; and 58.8 ± 18.4 vs 64.5 ± 24.1 mL, P < .01, respectively), whereas no significant difference was observed for ejection fraction (47.8 ± 8.5% vs 48.2 ± 10.8%, P = .57) ( Figure 5 ). The scatterplots in Figure 6 illustrate the moderate to high correlations between RT3DE imaging and MRI, with correlation coefficients ranging from 0.72 to 0.84 ( P < .001 for all correlation analyses). Mean differences as assessed by Bland-Altman analysis were 10.2 ± 21.2 mL (95% confidence interval [CI], −31.3 to 51.7 mL) for RV end-diastolic volume, 4.5 ± 14.5 mL (95% CI, −23.8 to 32.9 mL) for RV end-systolic volume, 5.7 ± 15.5 mL (95% CI, −24.6 to 36.0 mL) for RV stroke volume, and 0.4 ± 7.5% (95% CI, −14.2% to 15.1%) for RV ejection fraction ( Figure 6 ).
Intraobserver and Interobserver Variability
Intraobserver and interobserver variability for RTDE was assessed in 92 patients (92%). The mean time between measurements to determine the intraobserver variability was 154 ± 84 days. The results of linear regression analysis as well as the Bland-Altman plots are given in Figure 7 (intraobserver agreement) and Figure 8 (interobserver agreement). Tables 2 and 3 summarize, respectively, the intraobserver and interobserver measurements and their variability and intraclass correlation coefficients. The intraclass correlation coefficients ranged from 0.80 to 0.93 between measurements (performed by one observer) and from 0.76 to 0.95 between observers.