The clinical applicability and reliability of three-dimensional (3D) speckle-tracking echocardiography has not been well studied in pediatric patients. The aim of this study was to compare two-dimensional (2D) echocardiography and 3DE real-time full-volume–derived strain and rotation indices in healthy children and patients with dilated cardiomyopathy (DCM).
Children with either normal function or DCM were prospectively recruited in an outpatient setting, and deformation indices, including circumferential, radial, and longitudinal strain and torsion, were measured by 2D and 3D echocardiography. Control subjects were compared with patients using the Mann-Whitney U test, correlations between 2D and 3D measurements were analyzed using Spearman correlation analysis, and reproducibility analyses are reported using intraclass correlation coefficients and coefficient of variations.
The study cohort consisted of 15 patients (47%) with DCM and 17 control subjects (53%). The median age of the cohort was 13.4 years (range, 5.7–19.3 years). By both 2D and 3D analysis, magnitudes of global longitudinal ( P = .01), circumferential ( P = .007), and radial ( P = .004) strain were significantly lower in patients with DCM in comparison with control subjects. Using receiver operating characteristic curves, lower values of absolute circumferential (area under the curve = 0.95, P < .0001) and longitudinal (area under the curve = 0.93, P < .0001) strain were associated with left ventricular dysfunction. No difference was noted in torsion between control subjects and patients. Three-dimensional analysis was superior to 2D analysis in terms of intraobserver, interobserver, and test-retest reliability.
Left ventricular deformation shows significant changes while torsion is preserved in outpatients with DCM compared with control subjects. Three-dimensional global strain can discriminate subtle left ventricular dysfunction and has better reproducibility in comparison with 2D echocardiography. High-resolution 3D imaging is a useful clinical assessment tool for cardiac performance and may overcome some of the limitations of 2D analysis.
Three-dimensional speckle-tracking echocardiography-derived mechanics is feasible and a highly reproducible modality that can be used to characterize ventricular function.
This prospective study compared 3D and 2D echocardiography-derived mechanics in healthy children and in pediatric patients with dilated cardiomyopathy.
Global strain showed significant changes while torsion was preserved in outpatients with dilated cardiomyopathy compared to controls.
Three-dimensional echocardiography-derived mechanics is a useful clinical assessment tool for cardiac performance and may overcome some of the limitations of 2D analysis.
The evaluation of left ventricular (LV) mechanics using strain, twist, and torsion analyses has drawn significant attention in the recent years. Although analysis of strain on the basis of two-dimensional (2D) echocardiography has been shown to be superior to conventional measures of ventricular function in several clinical settings, a newly developed three-dimensional (3D) echocardiographic modality is useful in overcoming some of the inherent limitations of 2D echocardiography (out-of-plane motion). Two-dimensional echocardiographic analysis of strain may underestimate the impact of multidirectional axes of cardiac motion, which can reduce its accuracy. Recent studies have also proposed that 2D echocardiographic rotation analyses are not reliable.
Three-dimensional echocardiography using speckle-tracking echocardiography allows the analysis of multidirectional components of myocardial mechanics. In contrast to earlier methods of 3D acquisition, which consisted of multiple-beat subvolume reconstruction into a single virtual volume, recent improvements in technology have enabled single-beat, real-time full-volume (RTFV) 3D image capture. In addition, recent platforms have been optimized for higher volume rates of image acquisition. The development of knowledge-based automated analysis has improved the poor reproducibility that has been reported with 2D echocardiographic studies. Thus, RTFV 3D echocardiography shows promise as offering a more accurate representation of ventricular mechanics, and its applicability in disease states is an area of growing interest.
Some studies have implied the existence of age-related changes in mechanical indices, and others have shown varying data when comparing 2D and 3D echocardiographic deformation analyses. To further understand these relationships, in this first prospective study in the pediatric population, we compared both the magnitudes and the reliability of 2D and 3D echocardiographic RTFV-derived strain and rotation indices in healthy children and in patients with dilated cardiomyopathy (DCM).
Subjects 5 to 18 years of age who underwent clinically indicated echocardiographic examinations were prospectively recruited in an outpatient setting. Control subjects with normal intracardiac anatomy and function and study patients with diagnoses of DCM were enrolled after informed consent was obtained. Patients who were pacemaker dependent, and those with underlying structural lesions were excluded. This study was approved by the Stanford Medical School Committee on Clinical Investigation.
Images were acquired using a standard protocol by two cardiologists (T.A.T. and S.B.) with expertise in both 2D and 3D echocardiographic imaging, using the Acuson SC2000 ultrasound system (Siemens Medical Solutions, Mountain View, CA).
Two-dimensional images were acquired using 8V3c or 4V1c transducers with the subject in the lateral decubitus position. For each view, images were acquired over two or three heartbeats, and one cardiac cycle was analyzed. Machine settings were selected to provide sharp endocardial definition and a frame rate of approximately 80 Hz. Images were acquired in the apical four-chamber view focused for visualization of the left ventricle and the parasternal short-axis view at the mitral valve, mid papillary muscle, and LV apex. Ejection fraction was calculated using the 5/6 area-length method.
Three-dimensional RTFV data sets were obtained using the 4Z1c transducer for each patient. To ensure the inclusion of the entire left ventricle, orientation was adjusted to visualize the LV apex in both the four-chamber and two-chamber planes. All 3D echocardiographic images were full-volume reconstructed from a single beat. Image depth and width were adjusted to achieve a volume rate of approximately 40 volumes/sec.
LV Mechanics Analysis Using 2D Echocardiography
Postprocessing of 2D echocardiographic images was performed using the eSie VVI technology 2D speckle-tracking software version 3.0 (Siemens Medical Solutions). The endocardial and epicardial layers were manually traced. Global strain analysis included the standard 16-segment model, with the global values derived from the average of all segments, and the peak systolic strain values were recorded. Circumferential and radial strain were acquired at the mid papillary muscle level, and longitudinal strain was analyzed from the four-chamber view. LV torsion (degrees per centimeter) was calculated as the difference in rotation (degrees) between the LV apex and the base, divided by the end-diastolic length of the left ventricle measured from the four-chamber view.
LV Mechanics Analysis Using 3D Echocardiography
Postprocessing of 3DE images was performed with a dedicated voxel (3D) tracking software with automated ventricular contouring capabilities, previously described. The new application, eSie Mechanics, leverages the knowledge-based, expert-trained, automated contouring solution provided whereby the endocardium is detected in the first frame of the heart cycle. The left ventricle is automatically segmented. Deformation was analyzed in the circumferential, longitudinal, and radial directions; global strain and rotation indices were analyzed. Two or three cardiac cycle clips were acquired, and then all analysis results were averaged. A representative analysis of a volume set and the related output is shown in Figure 1 .
The paired t test was used to compare differences in 2D and 3D measurements. The nonparametric Mann-Whitney U test was used to compare differences in indices of mechanics between patients with DCM and control subjects. Correlations between ventricular ejection fraction and deformation indices and between 2D and 3D deformation indices were analyzed using the Spearman correlation coefficient ( r ).
Receiver operating characteristic curves were generated to evaluate strain indices as potential measures associated with LV dysfunction defined as an LV ejection fraction < 55%.
At least two sets of images were obtained in 10 subjects, which included both control subjects and patients. Intra- and interobserver variability was evaluated on a single volume set obtained and analyzed by two investigators (S.B. and S.N.S.). Test-retest analysis was performed by obtaining and analyzing two separate volume sets on a single patient by two independent investigators (S.B. and T.A.T.), allowing complete independence of these measures. Volume sets for test-retest reliability were obtained using the same equipment and location and within a short period of time (<5 min). All investigators were blinded to the results of the second data set. Reproducibility results are reported as intraclass correlation coefficients and coefficients of variation (calculated as the absolute difference divided by the mean of the repeated observations and expressed as percentages).
The baseline clinical characteristics are summarized in Table 1 . Of the 32 patients recruited, 15 had histories of DCM, with varying degrees of remodeling, and were being followed in an outpatient setting. Of these patients, the etiologies of DCM included familial ( n = 5), metabolic ( n = 1), infectious ( n = 1), LV noncompaction ( n = 3), and other ( n = 5). Three of the 15 patients with DCM had recovered ventricular function at the time of recruitment.
|Control subjects||Patients with DCM|
|n||17 (53)||15 (47)|
|Male||8 (47)||10 (67)|
|Age at study (y)||13.9 (8.5 to 19.3)||13.3 (5.7 to 19.3)|
|Ejection fraction (%)||65 (55 to 75)||47 (20 to 67)|
|LVEDD (cm)||4.67 (3.86 to 5.85)||5.6 (3.7 to 7.4)|
|Median LVEDD Z score||−0.07 (−1.0 to +1.26)||+2.9 (−1.7 to +7.4)|
|NYHA class I or II||17 (100)||14 (93)|
|NYHA class III or IV||1 (7)|
Comparison of Cardiac Mechanics Analysis Using 2D and 3D Speckle-Tracking Echocardiography
Two-dimensional echocardiographic frame rates were significantly higher than the volume rates achieved with 3D echocardiography ( Table 2 ). Despite this observation, 2D and 3D strain analysis produced similar differences when comparing healthy control subjects and patients with DCM ( Table 2 , Figures 1 A–1C).
|Control subjects||Patients with DCM||P|
|Frame rate (Hz)||81 (56 to 104)||76 (56 to 97)||NS|
|Ratio of frame rate (Hz) to heart rate (beats/min)||1.2 (0.7 to 1.7)||1.0 (0.7 to 1.9)||NS|
|Twist (°)||5.7 (2.9 to 14.9)||5.4 (0.9 to 11.9)||NS|
|Torsion (°/cm)||1.1 (0.7 to 2.5)||0.8 (0.2 to 2.1)||NS|
|Circumferential strain (%)||−25.2 (−39.1 to −20.2)||−18.8 (−26.6 to −5.7)||<.0001|
|Longitudinal strain (%)||−21.3 (−24.2 to −13.2)||−12 (−19.7 to −5.2)||<.0001|
|Radial strain (%)||35.9 (17.8 to 53.4)||29.8 (7.6 to 62.7)||NS|
|Volume rate (volumes/sec)||41 (24 to 54)||43 (21 to 62)||NS|
|Ratio of volume rate (volumes/sec) to heart rate (beats/min)||0.6 (0.3 to 0.9)||0.5 (0.4 to 1.2)||NS|
|Twist (°)||14.6 (12.6 to 18.4)||14.2 (12.5 to 19.2)||NS|
|Torsion (°/cm)||2.4 (1.7 to 4.8)||2.1 (1.7 to 4.2)||NS|
|Circumferential strain (%)||−26.9 (−40.3 to −20.3)||−22.8 (−31.2 to −13.9)||<.01|
|Longitudinal strain (%)||−21.1 (−28.8 to −15.9)||−14.8 (−24.1 to −10.0)||<.05|
|Radial strain (%)||34.3 (22.4 to 48.8)||25.2 (11.6 to 41.5)||<.01|
In contrast, notable differences in the magnitudes of twist and torsion indices were present between 2D and 3D echocardiography. By 3D analysis, both parameters were significantly higher in comparison with 2D analysis ( Table 2 , Figure 2 ). This magnitude of difference was preserved both in healthy control subjects and in patients with DCM ( Figure 2 ).
Two-dimensional and 3D global longitudinal and circumferential strain exhibited a fair correlation, whereas radial strain and torsion did not ( Figure 3 ).
Distinguishing between Control Subjects and Patients with DCM
As expected, there were significant differences in ejection fraction, global longitudinal strain ( Figure 1 A), and circumferential strain ( Figure 1 B) between healthy control subjects and patients with DCM in both the 2D and 3D analyses. In contrast, radial strain was not different between control subjects and patients with DCM in the 2D analysis ( Figure 1 C). Furthermore, no appreciable differences were noted in global twist and torsion ( Figure 2 ) values between the control and disease groups. A separate analysis that dichotomized the cohort by manifest systolic dysfunction using an ejection fraction < 55% also did not demonstrate appreciable difference in twist or torsion. However, significant differences were seen in global longitudinal, circumferential, and radial strain in both 2D and 3D analyses (data not shown).
Global circumferential, longitudinal, and radial strain were correlated with ventricular systolic function as measured by ejection fraction ( Figures 4 A–4C). This correlation was present in both 2D and 3D analysis. In contrast, no significant association was noted between torsion and twist indices and ventricular systolic function either by 2D or 3D analysis ( Figure 5 ).
Receiver operating characteristic curve analysis demonstrated strong associations between 3D global circumferential (area under the curve [AUC] = 0.87, P = .001) and longitudinal strain (AUC = 0.87, P < .001) and LV dysfunction. Similar associations were seen between LV dysfunction and 2D-derived circumferential strain (AUC = 0.95, P < .0001) and longitudinal strain (AUC = 0.93, P < .0001).
Reproducibility of Mechanics Derived from 2D versus 3D Echocardiography
Intraobserver, interobserver, and test-retest reliability were all superior for 3D versus 2D analysis, with an average intraclass correlation coefficient of 0.93 and a coefficient of variation <10% for 3D echocardiography compared with 0.71 and 10% to 48%, respectively, for 2D echocardiography.
Comparison of reproducibility for 2D and 3D analyses is summarized in Table 3 . With the exception of 3D circumferential strain, all other comparisons yielded excellent test-retest reliability and coefficients of variation <15%.
( n = 10)
( n = 10)
( n = 10)
|ICC||CoV (%)||ICC||CoV (%)||ICC||CoV (%)|
|Circumferential strain||0.70 (−0.10 to 0.92)||19.4||0.69 (−0.78 to 0.92)||26.0||0.94 (0.77 to 0.99)||10.9|
|Longitudinal strain||0.86 (0.46 to 0.96)||17.8||0.93 (0.74 to 0.98)||10.4||0.92 (0.68 to 0.98)||12.8|
|Radial strain||0.64 (−0.23 to 0.91)||21.9||−0.42 (−7.9 to 0.68)||31.5||0.34 (−0.45 to 0.72)||41.9|
|Apex-base torsion||0.60 (−0.30 to 0.89)||48.3||0.80 (0.26 to 0.95)||30.0||0.72 (0.13 to 0.93)||31.9|
|Circumferential strain||0.99 (0.96 to 0.998)||6.6||0.98 (0.49 to 0.997)||7.0||0.45 (−0.38 to 0.76)||13.6|
|Longitudinal strain||1.00 (0.98 to 0.999)||2.6||0.99 (0.96 to 0.998)||3.0||0.93 (0.72 to 0.98)||9.9|
|Radial strain||0.97 (0.87 to 0.99)||6.6||0.98 (0.77 to 0.995)||7.0||0.75 (0.02 to 0.94)||13.4|
|Apex-base torsion||1.00 (0.98 to 0.999)||3.5||0.98 (0.93 to 0.995)||4.0||0.84 (0.36 to 0.96)||15.0|