To evaluate the accuracy of real-time three-dimensional echocardiography (RT3DE) using a biplane and multiplane method in determining left ventricular (LV) mass compared to cardiac magnetic resonance imaging (CMR).
LV mass was measured in 18 adult patients with congenital aortic stenosis using CMR and echocardiography (M-mode, two-dimensional echocardiography (2DE), and RT3DE). RT3DE data were analysed using a biplane and multiplane method. No geometric assumptions were necessary using the multiplane RT3DE method.
With regard to biplane or multiplane RT3DE, no tendency of over- or underestimation of LV mass was observed. Pearson's correlation coefficients for RT3DE versus CMR were 0.84 and 0.90 for the biplane and multiplane method, respectively. In addition, the accuracy of both RT3DE methods were comparable (Fisher's R -to- Z transformation: Z = 0.69, P = NS). Finally, off-line analysis using biplane RT3DE was significantly faster than multiplane RT3DE (3.8 ± 1.2 vs. 7.8 ± 1.7 minutes, P < 0.001).
Biplane RT3DE provided an accurate estimate of LV mass in patients with concentric left ventricular hypertrophy, which was not improved by multiplane RT3DE. The accuracy and speed of analysis renders biplane RT3DE an attractive tool in daily clinical practice for assessing the degree of LV hypertrophy.
Left ventricular (LV) hypertrophy is associated with increased morbidity and mortality in patients with congenital aortic stenosis and may be an important risk factor for long term outcome. 1 To use the severity of LV hypertrophy as an indicator for surgical intervention, a reliable, reproducible, and widely available diagnostic method is warranted. Echocardiography is the imaging modality most commonly used to evaluate LV mass because of its practical and economic advantages over cardiac magnetic resonance (CMR). However, the accuracy of linear (M-mode) or 2-dimensional echocardiography (2DE) images is limited, as they both rely on geometric assumptions of uniform chamber size and shape. Furthermore, the unintended use of oblique planes leads to overestimation of LV mass measurements by M-mode, and foreshortening of apical views leads to underestimation of LV mass measurements by 2DE. 2–7
Considering these limitations, real-time three-dimensional echocardiography (RT3DE) appears to be an attractive alternative for accurate evaluation of LV mass, especially since it can be measured directly from the three-dimensional data sets. Several methods for LV mass analysis from RT3DE data sets are available. A limited approach of RT3DE data is to select the nonforeshortened 2- and 4-chamber long-axis view and compute LV mass derived from model based biplane volumes (biplane RT3DE analysis). 6 A more comprehensive approach is to trace the endocardial and epicardial surfaces in multiple long-axis planes and correct the tracings in the short-axis views (multiplane RT3DE analysis). We hypothesize that a more accurate estimate of LV mass will be acquired with multiplane RT3DE analysis, as no geometrical assumptions are required. 8–10
The aim of the present study was to evaluate the accuracy of biplane and multiplane RT3DE compared to the current ‘gold standard’ for LV mass values obtained from CMR images. In addition, we evaluated if RT3DE was more accurate than the conventional M-mode and 2DE techniques.
We prospectively recruited 18 consecutive adult patients (13 men, age 30 ± 8 years, weight 80 ± 12 kg) with congenital aortic stenosis who visited the outpatient clinic for Adult Congenital Heart Disease of the Thoraxcentre. The M-mode, 2DE and RT3DE data acquisition were performed on the same day, and the CMR study was performed 26 ± 14 days after the echocardiographic data acquisition. There were no changes in the patient's clinical status between the echocardiographic examination and the CMR study. The mean peak aortic jet velocity was 3.7 ± 0.8 m/s and all patients were in sinus rhythm. The institutional review board approved the study, and all patients gave written informed consent.
A clinical 1.5-tesla MRI scanner with a dedicated cardiac four-element phased-array receiver coil was used for imaging (Signa CV/I, GE Medical Systems, Milwaukee, Wisconsin USA). Repeated breath-holds and gating to the electrocardiogram were applied to minimize the influence of cardiac and respiratory motion on data collection. Cine-MRI was performed with a steady-state free-precession technique (FIESTA, GE) with the following imaging parameters: repetition time, 3.0 to 3.4 ms; echo time, 1.5 ms; flip angle, 45°; temporal resolution, 47 ms; section thickness, 8 mm; section gap, 2 mm; field of view, 300 × 340 mm; and matrix size 224 × 256. To cover the entire LV, 10 to 12 consecutive slices were acquired in short axis orientation and a single 4-chamber and 2-chamber orientation.
CMR cine loops were analysed offline with dedicated commercial software (CAAS-MRV, Pie Medical Imaging, Maastricht, The Netherlands) by one experienced CMR cardiologist (RJMvG) blinded to the results of the echocardiographic analysis. The endocardial and epicardial border of the LV were traced manually on the long-axis views. The calculated intersection points with the short-axis images were the basis of the automatic contour detection. 11 In every short-axis slice, manual corrections were performed when necessary, including the papillary muscles and trabeculations in the LV cavity. Myocardial mass was calculated by taking the difference of the end-diastolic endocardial and epicardial volume multiplied by 1.05 g/mL, which is the density of myocardium. 12 , 13
Data acquisition was performed by one experienced sonographer (JSM) using second harmonic imaging with a matrix-array transducer (×4, 2 to 4 MHz) connected to a commercial ultrasound system (SONOS 7500, Philips Medical Systems, Best, The Netherlands). Care was taken to include the entire LV within the pyramidal scan volume. RT3DE datasets were then acquired with wide-angled acquisition (93° × 84°) mode in which 4 wedge-shaped sub-volumes (93° × 21° each) were obtained from 4 different cardiac cycles during held respiration without moving the transducer. Acquisition was triggered to the R wave of every other cardiac cycle to allow time for storage of each sub-volume, resulting in a total acquisition time of 7 heartbeats.
The RT3DE datasets were analysed using 2 different commercial software packages: 3DQ-Qlab (Philips Medical Systems, Best, The Netherlands) and 4D Echo-View (TomTec Imaging Systems GmbH, Munich, Germany) for the biplane and multiplane analysis, respectively. Two independent observers (SCY and AN) blinded to the results of the CMR analysed the RT3DE datasets by using first the biplane method, followed by the multiplane method. LV mass was calculated with the use of a specific mass of myocardial tissue of 1.05 g/mL.
Biplane RT3DE data analysis
The pyramidal volume data were displayed in 3 different cross-sections ( Figure 1 A ). The anatomically correct 2- and 4-chamber views with the largest long-axis dimensions were selected as previously described by Mor-Avi et al.6 In these 2 planes, endocardial and epicardial contours were traced manually at end-diastole. The papillary muscles and trabeculations were included in the LV cavity to be consistent with the CMR measurements. The traced contours were then used to calculate LV volumes by use of the biplane Simpson formula incorporated in the analysis software. In addition, the LV long-axis dimension in the 2- and 4-chamber view was measured as the distance between the level of the mitral annulus and the most distal point at the apical endocardium.
Multiplane analysis RT3DE data
Multiplane analysis was performed as previously described by our group 8 . The anatomically correct 2- and 4-chamber views with the largest long-axis dimension were selected in the same manner as described above ( Figure 1 B ). Around this user-defined LV long axis, the software generated 8 uniformly spaced apical long-axis images 22.5° apart. In each view, epicardial and endocardial contours were manually traced at end-diastole with the papillary muscles and trabeculations included in the LV cavity. Once the points were manually traced on the eight long-axis planes, the position of the manually traced contours was verified in multiple short axis views from base to apex, and corrected when necessary ( Figure 2 ). The traced contours were then used to calculate LV myocardial volume and mass.
2D Echocardiographic measurements
The 2DE imaging was performed by one experienced sonographer (JSM) with a S3 transducer with second harmonic imaging from the apical window with the patient in the left lateral decubitus position. The 4- and 2-chamber views were acquired during held respiration, while care was taken to avoid foreshortening. Images were stored digitally and analysed offline (EnConcert, Philips Medical Systems) by 2 independent observers (SCY and AN). For both apical views, end-diastolic frames were selected at the start of the R wave. In each view, endocardial and epicardial contours, including the papillary muscles and trabeculations in the LV cavity, were traced manually. The traced contours were used to calculate endocardial and epicardial LV volumes by the biplane Simpson's formula. 14 The difference between the epicardial and endocardial volumes was computed for each view and multiplied by the specific mass of myocardial tissue to represent a biplane estimate of LV mass. In addition, the LV long-axis dimension in the 2- and 4-chamber view was measured as described for biplane RT3DE.
M-mode imaging was performed by one experienced sonographer (JSM) from a parasternal long-axis position using a standard transthoracic transducer. Measurements of the interventricular wall thickness (IVS), posterior wall thickness (PW), and LV internal diameter (LVID) were performed at end-diastole according to the recommendations of the American Society of Echocardiography. 15 LV mass was calculated according to the cube formula using the correction described by Devereux et al. : LV mass (g) = 0.8 [1.04((IVS + LVID + PW) 3 – LVEDD 3 )] + 0.6. 16
Inter- and intra-observer variability
To determine the interobserver variability for 2DE and RT3DE evaluations of LV mass, all measurements were repeated by a second observer (AN) blinded to the values obtained by the first observer (SCY). To assess intraobserver variability, all measurements were repeated 1 month later by an observer (SCY) blinded to the results of the previous measurements. Inter- and intraobserver variability was calculated as the absolute difference between the 2 measurements in percent of their mean. The inter- and intraobserver variability could not be measured for the M-mode technique as the measurements were already performed during acquisition, making blinding impossible.
All values were expressed as mean ± SD. The difference between each echocardiographic technique and CMR was evaluated by a paired t test. Agreement between techniques was evaluated by linear regression analysis with Pearson's correlation coefficient and the standard error of the estimate (SEE). The difference between the correlation coefficients of the biplane and multiplane RT3DE was tested using the Fisher R -to- Z transformation test. In addition, Bland-Altman analysis was used to determine the bias and 95% limits of agreement (1.96 SD) between echocardiographic measurements and CMR. 17 These analyses were performed for the first observer (SCY). Inter- and intraobserver variability values were averaged for all patients and tested by use of a paired t test for significance of differences between techniques. Values of P <0.05 were considered significant.
Acquisition of RT3DE data sets was feasible in all patients (acquisition time <2 minutes). The CMR value of LV mass was 177 ± 48 g. Off-line image processing and tracing required 2.0 ± 0.4 minutes for 2DE, 3.8 ± 1.2 minutes for biplane RT3DE and 7.8 ± 1.7 minutes for multiplane RT3DE ( P < 0.001 between techniques). Linear regression analysis and Bland-Altman analysis are shown in Figures 3 and 4 , respectively.
Mean LV mass with M-mode echocardiography was 209 ± 68 g, which was significantly higher than the CMR values ( P = 0.005). A Pearson's correlation of r = 0.79 (SEE 43 g, P < 0.001) with CMR values was observed. Bland-Altman analysis confirmed the overestimation by M-mode echocardiography by demonstrating a bias of 32 g with 95% limits of agreement at ± 82 g.
Mean LV mass with 2DE was 155 ± 31 g, which was significantly lower than the CMR values ( P = 0.02). A correlation of r = 0.69 (SEE 23 g, P < 0.001) with CMR values was observed. Bland-Altman analysis confirmed the underestimation by 2DE by demonstrating a bias of −22 g with 95% limits of agreement at ± 68 g. The LV long-axis dimension was 8.8 ± 0.8 cm in the 4-chamber view and 8.8 ± 0.6 cm in the 2-chamber view.
Biplane RT3DE yielded a mean LV mass of 183 ± 46 g, which was similar to the CMR values ( P = NS). A Pearson's correlation of r = 0.84 (SEE 26 g, P < 0.001) with CMR values was observed. Bland-Altman analysis showed no significant over- or underestimation by the RT3DE-technique as reflected by a bias of 6 g with 95% limits of agreement at ± 53 g. The LV long-axis dimension was 9.2 ± 0.9 cm in the 4-chamber view and 8.8 ± 0.7 cm in the 2-chamber view. The LV long-axis value in the 4-chamber view was significantly larger using biplane RT3DE compared to 2DE ( P < 0.05).
Finally, multiplane RT3DE yielded an LV mass of 181 ± 41 g, which was similar to the CMR values ( P = NS). A correlation of r = 0.90 (SEE 20 g, P < 0.001) with CMR values was observed. Bland-Altman analysis showed no significant over- or underestimation by the RT3DE-technique as reflected by a bias of 4 g with 95% limits of agreement at ± 41 g. The accuracy of LV mass measurement was similar between the biplane and multiplane RT3DE method (Fisher's R -to- Z transformation: Z = 0.69, P = NS).
The interobserver variability was 17 ± 13% for the 2DE technique, 9 ± 7% for the biplane RT3DE technique, and 11 ± 6% for the multiplane RT3DE technique ( P < 0.05 between 2DE and biplane RT3DE, P = 0.11 between 2DE and multiplane RT3DE). No difference in interobserver variability was observed between the biplane and multiplane RT3DE technique. The intraobserver variability was 8 ± 6% for the 2DE technique, 9 ± 9% for the biplane RT3DE technique, and 11 ± 7% for the multiplane RT3DE technique ( P = NS between techniques).
As expected, LV mass measurements from RT3DE data could be achieved with higher accuracy and lower interobserver variability than conventional echocardiographic techniques (M-mode, 2DE). Interestingly, LV mass measurement by biplane RT3DE was as accurate as multiplane RT3DE in patients with concentric LV hypertrophy.
Quantification of LV mass has traditionally been based on M-mode measurements of myocardial thickness, coupled with geometrical modelling of the LV, or model-based calculations from manually traced endocardial and epicardial contours obtained from 2DE images. Our results, however, showed that M-mode significantly overestimates LV mass probably due to oblique cuts, and that 2DE underestimates LV mass due to foreshortening. 2–7 An improved accuracy of LV mass measurement was found for RT3DE, either by biplane or multiplane analysis, compared to conventional echo techniques. Measurements of the LV long-axis dimensions showed that offline cross sectioning of the RT3DE data provided significantly less foreshortened and thus anatomically more correct apical views than conventional 2DE measurements, thus explaining the improved accuracy of biplane RT3DE analysis. Furthermore, we hypothesized that the accuracy of RT3DE could be improved further by increasing the number of long-axis planes used to trace the endocardial and epicardial boundaries and to correct the tracings on short-axis images. We found, however, that the accuracy of multiplane RT3DE analysis was not better than biplane RT3DE analysis using 2 orthogonal planes. A similar outcome has been shown by Teupe et al. , reinforcing the fact that using multiple planes does not increase accuracy in normally shaped ventricles. 18 Multiplane RT3DE analysis may be of extra value in patients with nonconcentric LV hypertrophy, like hypertrophic obstructive cardiomyopathy, but not convincingly in a population with concentric LV hypertrophy. 9 We recently demonstrated that LV mass measurement can be achieved with high accuracy by multiplane RT3DE in patients with abnormally shaped left ventricles due to congenital heart disease. 8
Another important finding of our study is that RT3DE measurements varied less between observers compared to 2DE measurements (significant for the biplane analysis, and showing a trend for multiplane analysis), reflecting reduced operator dependency. Both the interobserver and intraobserver variability of RT3DE are ± 10%, indicating that this technique is acceptable for clinical use and will improve the reliability of follow-up data. CMR, however, remains superior with variability values 2–4% reported by multiple investigators. 13 , 19 This difference may be explained by the relatively limited ability of ultrasound imaging compared with CMR to visualize the endocardial and epicardial borders.
The time needed for the acquisition and analysis of RT3DE data are acceptable. The acquisition time for one RT3DE data set is relatively fast and is limited to 7 heartbeats, with an acquisition time of less than 2 minutes. Off-line image processing and tracing required 3.8 ± 1.2 minutes for biplane RT3DE analysis and 7.8 ± 1.7 minutes for multiplane RT3DE analysis. Interestingly, it is possible to directly measure LV mass using the biplane RT3DE method on the ultrasound system (not in this study) during acquisition thus avoiding off-line analysis. Recently, custom software has been developed, incorporating semi-automated detection of the LV endocardial and epicardial surface, which will allow even more rapid and accurate measurement of LV mass. 20
Several limitations are inherent to RT3DE. The size of the heart may be too big to fit into the scan volume thereby reducing accurate measurements of LV mass. 9 Fortunately, this was not the case in our population, but this can be encountered in patients with LV dilatation or ventricular aneurysms. Another limitation of RT3DE is that factors that reduce 2DE image quality, including a small intercostal space and abundant body fat, also apply for RT3DE. In addition, the relatively large footprint of the ×4 matrix transducer probably influences image quality in a negative way. Technical improvements in the near future will probably include an expansion of the scan volume without loss of spatial resolution and a smaller footprint of the matrix transducer.
Several potential limitations must be noted. Test–retest variability (summation of variability in acquisition and analysis), which is more representative for the clinical situation, was not tested in the current study. Despite this shortcoming, one would expect that the variability in acquisition is lower when using RT3DE as it encompasses the whole left ventricle in one dataset in contrast to 2DE. This should however be confirmed in another study. Together with the given sample size, all conclusions of the present study must be drawn with caution.
Despite several limitations of RT3DE, our study showed that RT3DE data obtained with commercially available equipment provided images with sufficient detail to allow easy and fast offline analysis of LV mass. The biplane RT3DE method is a rapid and accurate method for clinical assessment of LV hypertrophy, avoiding overestimation of LV mass using M-mode which is currently the most applied clinical method for determining LV hypertrophy. Multiplane RT3DE might be the preferred method for eccentric LV hypertrophy and abnormal shaped ventricles. Biplane RT3DE is as accurate as multiplane RT3DE in patients with concentric LV hypertrophy. Although CMR remains the gold standard for assessement of LV hypertrophy, in daily clinical practice biplane RT3DE is an attractive tool due to the low costs, easy accessibility, and speed of acquisition and analysis.
Conflicts of interest: none declared.