Real-Time 3-Dimensional Echocardiography

Abstract

HomeCirculationVol. 119, No. 2Real-Time 3-Dimensional Echocardiography Free AccessReview ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toFree AccessReview ArticlePDF/EPUBReal-Time 3-Dimensional EchocardiographyAn Integral Component of the Routine Echocardiographic Examination in Adult Patients? Victor Mor-Avi, PhD, Lissa Sugeng, MD and Roberto M. Lang, MD Victor Mor-AviVictor Mor-Avi From the University of Chicago Medical Center, Chicago, Ill. Search for more papers by this author , Lissa SugengLissa Sugeng From the University of Chicago Medical Center, Chicago, Ill. Search for more papers by this author and Roberto M. LangRoberto M. Lang From the University of Chicago Medical Center, Chicago, Ill. Search for more papers by this author Originally published20 Jan 2009https://doi.org/10.1161/CIRCULATIONAHA.107.751354Circulation. 2009;119:314–329One of the most significant developments of the last decade in ultrasound imaging of the heart was the evolution of 3-dimensional (3D) imaging from slow and labor-intense offline reconstruction to real-time volumetric imaging. This imaging modality provides valuable clinical information that empowers echocardiographers with new levels of confidence in the diagnosis of heart disease. We have previously described the technological milestones in the development of real-time 3D echocardiographic (RT3DE) imaging and its major advantages over conventional 2-dimensional echocardiography (2DE) and reviewed the published literature that supported the use of this new methodology in clinical practice.1 Since 2006, the growing availability of RT3DE technology, its ease of use, and its multiple attractive features have sparked significant interest in the research community, resulting in a large number of publications, most of which have endorsed RT3DE imaging for clinical use by demonstrating its unique capabilities in different scenarios. In parallel, the clinical acceptance of this new tool has broadened significantly. The most recent clinically significant addition is matrix-array transesophageal echocardiography (TEE), which provides images of unprecedented quality that aid surgeons and interventional cardiologists in planning and guiding procedures and evaluating their outcomes.The purpose of the present article is to review the most recent RT3DE literature and provide readers with an update on the latest developments and the current status of this noninvasive imaging tool. Because different potential applications of RT3DE imaging have been explored to various extents, they are described here separately, and each is discussed with an emphasis on the scientifically established facts, along with the known stumbling blocks and difficulties.Quantification of LV Volumes and FunctionA firmly established advantage of 3D imaging over cross-sectional slices of the heart is the improvement in the accuracy of the evaluation of left ventricular (LV) volumes and ejection fraction (EF) by eliminating the need for geometric modeling, which is inaccurate in the presence of aneurysms, asymmetrical ventricles, or wall motion abnormalities, and the errors caused by foreshortened views even in symmetrical ventricles. The value of RT3DE imaging in this context has been demonstrated by multiple studies that compared RT3DE volume measurements with widely accepted reference techniques, including radionuclide ventriculography and cardiac magnetic resonance (CMR).2–6 These studies and others have demonstrated higher levels of agreement between the RT3DE approach and the respective reference technique compared with conventional 2DE methodology. Additionally, RT3DE measurements were found to be more reproducible than 2DE4,5,7 and, in some studies, even as reproducible as CMR.8 Interestingly, in a recent study in patients after myocardial infarction (MI),9 RT3DE-derived LV volumes were more accurate and more reproducible than those obtained with 201Tl-gated single-photon emission computed tomography compared with a CMR reference.The improved accuracy and reproducibility of RT3DE-based LV volume and EF measurements are of vital importance because clinical decision making relies heavily on these measurements in multiple clinical scenarios. In addition, these findings translate into smaller numbers of patients required to test a hypothesis, promising to result in significant savings in future studies aimed at assessing the effects of new drugs. Indeed, this trend was demonstrated by a recent follow-up study in patients after MI in which, similar to CMR, serial RT3DE measurements had low test-retest variability and were thus able to detect with confidence subtle changes in LV volumes over time that were not detectable by 2DE.10 Similar findings were described in another study aimed at risk stratification in patients after MI and patients with heart failure.11 However, despite the high correlation with the CMR reference values and the high reproducibility, several studies have reported that RT3DE-derived LV volumes were significantly underestimated.7,8,10,12–16Different possible explanations have been offered that focused mostly on intertechnique acquisition and analysis differences, but none of the studies were able to conclusively identify the main sources of error. Importantly, the degree of underestimation varied widely between these single-center studies from a few milliliters to considerable biases as high as 30% of the measured values. One possible explanation for the variable degrees of underestimation is that RT3DE data sets were analyzed differently by different investigators. Indeed, 2 approaches are commonly used for LV quantification from RT3DE data sets (Figure 1). One approach is based on selecting from a pyramidal RT3DE data set 2 anatomically correct nonforeshortened 2D views from which LV volume is calculated using a biplane approximation,7,17 the same as that used with 2D imaging (Figure 1, left). Although this 3D-guided biplane technique can minimize LV foreshortening, it still relies on geometric modeling to calculate volumes and is thus likely to be inaccurate in distorted ventricles. In an attempt to minimize this problem, investigators used larger number of planes to interpolate LV endocardial surface16,18 with partial success. Another approach to quantify LV volumes from RT3DE data sets, which was recently implemented in commercial analysis software, is based on semiautomated detection of LV endocardial surface, followed by calculation of the volume inside this surface either for selected phases such as end systole and end diastole or throughout the cardiac cycle19 (Figure 1, right). Because this approach uses direct volumetric quantification, it is not affected by LV foreshortening and does not rely on geometric modeling; thus, not surprisingly, it was found to be more accurate5,7,16 regardless of wall motion abnormalities8 and distorted ventricular shape.16 Our study7 determined the magnitude of LV volume underestimation in such patients by each of these 2 techniques compared with a CMR reference (Figure 2). Download figureDownload PowerPointFigure 1. Comparison chart for 2 approaches to LV volume measurement from RT3DE data sets: 3D-guided biplane analysis (left) and direct volumetric analysis (right). See text for details.Download figureDownload PowerPointFigure 2. Biases compared with CMR reference values noted in end-diastolic (EDV) and end-systolic (ESV) volumes and EF in a group of 20 patients. Color bars represent conventional 2DE biplane measurements, 3D-guided biplane measurements obtained from RT3DE data sets, and direct volumetric measurements obtained by detecting LV endocardial surface. ✓ Indicates resolved; x, remains unresolved. Reproduced by permission of Oxford University Press from Jacobs LD et al7 (Rapid Online Quantification of Left Ventricular Volume From Real-Time Three-Dimensional Echocardiographic Data. Eur Heart J. 2006;27:460–468).Nevertheless, even direct volumetric analysis, which is the more accurate of the 2 techniques, was found to significantly underestimate LV volumes, threatening to undermine the usefulness of RT3DE evaluation of LV size and function. To investigate this issue in depth, we recently conducted a multicenter study designed to identify the potential sources of error and to determine their relative contributions to the underestimation of RT3DE-derived LV volumes.20 RT3DE and CMR reference data obtained in a large group of patients with a wide range of LV function were used to test several hypotheses that allowed us to resolve this controversy. RT3DE images were analyzed at 5 sites by experienced echocardiography investigators who had received various levels of instruction with prototype software for direct volumetric analysis (Figure 2, right) without being informed that the level of experience was one of the variables in the study design. Despite the high correlations with the CMR reference values at all sites (mostly >0.90), the biases progressively increased with decreasing level of specific experience, reaching 2 to 3 times those noted by the most experienced site.In the search for additional sources of error, CMR images were reformatted into 3D data sets and analyzed with the same software used for analysis of RT3DE date sets. These measurements resulted in virtually the same volume values and thus ruled out analysis-related differences as a significant source of error. Thereafter, several in vitro experiments were performed, including volume measurements in a phantom, which revealed that a barely visible 1-mm difference in the position of the endocardial surface resulted in a 10% difference in volume (Figure 3, top). Subsequent measurements in water-filled latex balloons showed that tracing along what appeared to be the water-latex interface resulted in volume measurements that were ≈10% below the true volume, whereas tracing further out, along the middle of the latex layer, yielded accurate volumes (Figure 3, bottom), even without the complicating effects of the papillary muscles and endocardial trabeculae. From these findings and the fact that biases in measured LV volumes in humans were related to image quality,6 we hypothesized that the spatial resolution of RT3DE images is not sufficiently high to differentiate between myocardial tissue and trabeculae (Figure 4, top) and that the more experienced investigators in the multicenter study traced the endocardium beyond the visible interface to compensate for this limitation, which resulted in larger LV volumes. To test this hypothesis, CMR images were reanalyzed with trabeculae excluded from the LV cavity (Figure 4, bottom). The use of these nonconventional reference values essentially eliminated the bias. Download figureDownload PowerPointFigure 3. Top, Long-axis cut plane of the egg-shaped phantom extracted from an RT3DE data set shown with the boundary traced along the interface (left), after expanding the boundary 1 mm outward (middle), and with both boundaries (right). Interestingly, the small difference between the 2 boundaries resulted in an 11% difference in the measured volume (V) of the 3D shell. Bottom, Cross-sectional views of a water-filled latex balloon with 3 alternative manually traced boundaries: along the inner interface (left), along the outer interface (middle), and in the center of the latex layer (right). Volumes resulting from each tracing session in this balloon are shown for comparison with the true volume of 150 mL. Modified with permission by Elsevier (Mor-Avi et al.20 Real-Time 3D Echocardiographic Quantification of Left Ventricular Volumes: Multicenter Study for Validation With Magnetic Resonance Imaging and Investigation of Sources of Error. J Am Coll Cardiol Imaging. 2008;1:413–423).Download figureDownload PowerPointFigure 4. Top, Example of short-axis cut planes extracted from RT3DE data sets. In 1 patient (left), trabeculae can be well visualized and clearly differentiated from the myocardium; in another patient (right), the spatial resolution of the RT3DE image is not sufficient to provide this kind of detail. Bottom, Example of a short-axis CMR slice with the endocardial surface traced to include trabeculae in the LV cavity (left) and, in a separate analysis, to exclude them (right) from the LV cavity. Modified with permission by Elsevier (Mor-Avi et al.20 Real-Time 3D Echocardiographic Quantification of Left Ventricular Volumes: Multicenter Study for Validation With Magnetic Resonance Imaging and Investigation of Sources of Error. J Am Coll Cardiol Imaging. 2008;1:413–423).The results of this multicenter study underscored the need for unified guidelines for tracing LV endocardial boundary to obtain RT3DE measurements of LV volumes comparable to the current standard reference CMR technique. Moreover, this study demonstrated that this reference technique relies on the user’s decision about which short-axis slices (basal and apical) should be included in LV volume calculation. This decision is subjective, depends on the choice of criteria, and significantly affects the reference values. In this regard, volumetric analysis based on detection of endocardial surface avoids these limitations and thus is more reproducible21 and potentially more accurate, assuming that the endocardial boundary can be well visualized, including clear differentiation between the myocardium and the trabeculae.In this respect, multiplane 2D imaging or 3D-guided triplane analysis may provide a practical solution for LV volume measurements in patients with suboptimal endocardial visualization in 3D data sets, which may be particularly useful in the setting of a busy clinical laboratory. This approach offers superior image quality because of improved line density while preserving the benefits of knowing the precise location and relationship of the image planes with respect to each other (Figure 5). Download figureDownload PowerPointFigure 5. Real-time triplane echocardiography from the apical window allows simultaneous visualization of apical 4-, 2-, and long-axis views (A4C, A2C, and ALAX, respectively). End-diastolic images were obtained at peak stress in a patient undergoing dobutamine stress test. The panel on the right shows the composite 3D display of all 3 views.RT3DE Evaluation of LV MassAs opposed to LV volume measurements, which require accurate identification of the endocardial boundaries, LV mass measurements rely on epicardial visualization, which is known to be even more challenging. This difficulty is in addition to inaccurate modeling and foreshortening. Nevertheless, in our initial studies aimed at RT3DE evaluation of LV mass by either the 3D-guided biplane technique17 or the volumetric analysis,22 the accuracy and reproducibility of RT3DE estimates were higher than those of the traditional M-mode and 2D techniques. More recently, these observations were confirmed without significant differences between techniques in a large group of patients with concentric LV hypertrophy.23 Moreover, volumetric measurements of LV mass were found to correlate highly with CMR reference values in patients with wall motion abnormalities8 and in patients with abnormally shaped ventricles secondary to congenital heart disease.24 Although the former study described a considerable negative bias, the latter reported only minimal biases. Similar to the inconsistencies with LV volume measurements, these differences are likely to be due to differences in strategies for identifying and tracing the endocardial and epicardial boundaries.RT3DE Assessment of Regional Wall MotionFurthermore, the ability of RT3DE imaging to almost instantaneously capture the entire heart into a data set that contains the complete dynamic information on LV chamber, from which the ventricle can be viewed in any arbitrary plane, suggested that RT3DE data sets are suitable for simultaneous analysis of regional wall motion in all LV segments. In a recent study,25 RT3DE-derived regional EF was validated against a CMR reference, and the feasibility of its use as an index of regional LV function was tested for objective detection of wall motion abnormalities. The methodology described in this study indicated that RT3DE-based analysis of regional LV function may also be potentially useful in different clinical applications, including stress testing and guidance of resynchronization therapy.Indeed, the fast volumetric imaging of the entire heart has indicated its potential usefulness in the context of exercise stress testing in which speed of acquisition of multiple views is crucial.26,27 Another study28 highlighted the advantages of using RT3DE imaging over 2DE during dobutamine stress testing (Figure 6). A more recent study29 demonstrated that RT3DE data sets contained sufficient information for the interpretation of dobutamine stress tests, which allowed accurate diagnosis of myocardial ischemia compared with single-photon emission computed tomography myocardial perfusion imaging, which was similar to conventional 2D methodology. However, an important advantage of the RT3DE approach, in addition to faster acquisition, was its ability to extract offline multiple short- and long-axis views of the ventricle that can help in determining the extent of wall motion abnormality and in ruling out artifacts frequently noted in standard imaging planes caused by limited endocardial visualization. Another study30 demonstrated high levels of agreement with 2DE-based wall motion scores. Download figureDownload PowerPointFigure 6. Offline viewing of real-time 3D data obtained during dobutamine stress test. These data sets can be used to extract multiple short-axis views at different levels of the left ventricle (top). Shown is an example of such views extracted from data sets obtained at rest (bottom left) and during peak dobutamine stress (bottom right).Contrast-Enhanced RT3DE ImagingRecently, LV triplane imaging with contrast enhancement was reported to be accurate and reproducible for the calculation of LV EF compared with CMR imaging.31 Although numerous previous studies have demonstrated contrast-related improvement in endocardial visualization on 2DE images, this does not automatically mean that similar improvements would be seen on RT3DE images. A probable reason is the increased microbubble destruction caused by an increase in ultrasound energy that is delivered into the entire scan volume during RT3DE imaging rather than into a thin slice, as in the case of 2D imaging. This problem can be minimized either by selective dual triggering at end systole and end diastole32 or by using low mechanical indexes with continuous harmonic imaging.15 One may speculate that alternative contrast-targeted imaging modes such as pulse inversion and power modulation may be useful in this context, but this remains to be tested in future studies.Contrast enhancement was found not only to improve the accuracy and reproducibility of LV volume measurements in patients with poor image quality but also to enhance the assessment of regional wall motion from RT3DE data sets. We found that with the use of selective dual triggering to minimize bubble destruction by ultrasound energy, contrast enhancement increased the accuracy of RT3DE-based analysis of regional LV function against CMR reference and its reproducibility to levels similar to those noted in patients with optimal image quality.33 Several other studies that used contrast enhancement with dobutamine stress have demonstrated that it improved segmental endocardial visualization34 and resulted in high levels of agreement with 2DE-based wall motion scores.35 Although the results of these studies may have important implications on the use of contrast for volumetric quantification of both global and regional LV function, no studies published to date have demonstrated the ability of contrast enhancement to improve the diagnostic accuracy of stress testing. Such studies are necessary for contrast-enhanced RT3DE imaging to become part of routine pharmacological stress testing.LV Shape AnalysisIn patients after MI or patients undergoing cardiac resynchronization therapy, evaluation of LV remodeling has traditionally been performed with 2DE-derived LV volumes. It is well known, however, that as LV function worsens, ventricular size increases and the ventricle becomes more globular than elliptical. Until now, these changes have been assessed with a 2D-derived sphericity index, which does not reflect discrete changes in regional LV shape. It has been postulated that RT3DE-based characterization of the LV endocardium may better reflect global and regional LV shape. Mannaerts and colleagues36 described a new 3DE-based sphericity index that was shown to be an earlier and more accurate predictor of LV remodeling in patients after an acute MI than other clinical ECG and echocardiographic parameters. The development of software for dynamic LV shape analysis from RT3DE data sets (Figure 7) promises to make the assessment of LV remodeling with this approach clinically useful.37Download figureDownload PowerPointFigure 7. Color-encoded parametric images of regional sphericity superimposed on 3D renderings of LV casts obtained in 2 patients: 1 with normal LV function (left) and 1 with dilated cardiomyopathy (right). Note the differences in EFs and the apical sphericity index (ASI) values (0.91 in the normal ventricle vs 0.28, reflecting LV shape changes in the dilated ventricle).RT3DE Evaluation of LV DyssynchronyTissue Doppler imaging (TDI) is currently considered the standard technique for selecting patients for cardiac resynchronization therapy because of its ability to quantify intraventricular dyssynchrony. Despite the excellent temporal resolution of TDI, this methodology has several limitations, including an inability to assess multiple myocardial segments in different planes simultaneously, an angle dependency that translates into the evaluation of the timing of longitudinal motion only, and an inability to reliably depict wall motion in the apical segments. In addition, despite the wealth of TDI-based dyssynchrony research, different investigators have used different approaches to quantify dyssynchrony, resulting in inconsistent conclusions. Although there is currently no accepted reference standard technique for LV dyssynchrony measurements, no TDI-based technique has been proved to reliably measure dyssynchrony in large clinical trials. Accordingly, other techniques for the quantification of intraventricular dyssynchrony are required.With its capability to quickly capture the 3D dynamics of the entire left ventricle, including the timing of regional wall motion independent of its direction, RT3DE imaging has emerged as an alternative approach for the quantification of LV dyssynchrony.38 Most RT3DE dyssynchrony studies have used the SD of the regional ejection times (interval between the R wave and minimum systolic LV volume) as an index of dyssynchrony. Recent attempts to compare this index against TDI have resulted in disparate results that ranged from fair intertechnique correlation39 to poor agreement, which was explained by the angle dependency of TDI40 and by the fact that these 2 techniques measure different parameters. RT3DE assessment of LV dyssynchrony has the potential advantage of measuring the timing of the longitudinal, radial, and circumferential motion as opposed to the longitudinal motion only, which is measured by TDI. Of note, the RT3DE dyssynchrony index was recently compared against phase analysis of gated single-photon emission computed tomography images41 and showed good intertechnique correlation.RT3DE was used in patients with heart failure, low EF, and wide QRS complex to predict the acute response to cardiac resynchronization therapy. It was shown that approximately two thirds of patients with high RT3DE-derived dyssynchrony index responded to biventricular pacing with a decrease in dyssynchrony and acute improvement in global LV function (Figure 8, left).42 Recent studies have shown a direct relationship between overall LV performance and synchronicity.43 In these studies, RT3DE assessment of LV dyssynchrony identified patients with heart failure, low EFs, and asynchronous LV contraction who could theoretically benefit from cardiac resynchronization therapy but would not be considered candidates because of their narrow QRS duration.43 Interestingly, contrary to expectations, biventricular pacing did not result in significant long-term benefits in this subgroup of patients with narrow QRS and increased intraventricular dyssynchrony as reflected by tissue Doppler parameters.44 One explanation for the failure of cardiac resynchronization therapy to improve long-term outcomes in approximately one third of the patients selected on the basis of established criteria could be that placement of the pacing electrodes was not optimized because of the lack of sufficiently accurate tools to guide the pacing catheter to the site of latest activation. RT3DE imaging also may prove useful in this regard because RT3DE-based mapping of the distribution of regional contraction times (Figure 8, right) shows which LV myocardial segments are the latest to contract.40,45 One of the emerging technological advancements in this context is the fusion of RT3DE-derived dyssynchrony maps with coronary vein computed tomography to guide the placement of the pacing catheter. Download figureDownload PowerPointFigure 8. Assessment of the improvement in synchrony of LV contraction with pacing. Regional volume-time curves (left) obtained in a patient with LV dyssynchrony without (top) and with (bottom) biventricular pacing. Endocardial surfaces reconstructed from each data set are shown with segmentation and color coding according to regional time to end ejection (middle) along with the bull’s-eye representation of the same data (right). Note the changes in colors with pacing reflecting the effects of resynchronization therapy in this parametric display.RT3DE Evaluation of Right Ventricular Volumes and FunctionThe ability to measure right ventricular (RV) volumes and function accurately is important in the management of congenital heart disease and primary pulmonary hypertension. Because of the complex geometrical crescent shape of this chamber, the estimation of RV volumes based on geometric modeling from 2D images has been extremely challenging. As a result, in clinical practice, tricuspid annular plane systolic excursion has traditionally been used as a surrogate for RV performance. Theoretically, the intrinsic ability of RT3DE imaging to directly measure RV volumes without the need for geometric modeling could be expected to result in improved accuracy and reproducibility compared with traditional 2DE measurements. Surprisingly however, the first study to compare these 2 techniques side by side against a CMR reference found that RT3DE measurements offered no significant advantage.47 A subsequent RT3DE study48 reported only slightly better levels of agreement with CMR. Similar to LV volume measurements, there are several ways to explain these findings, but the major sources of errors, which may be different for the 2 chambers, have not yet been identified for the right ventricle. Indeed, RV volume measurements were affected by multiple factors, including gain settings and thickness and orientation of disks used by the disk summation technique.49,50Another potential source of intermodality discordance is that the complex 3D shape of the right ventricle may affect the ability of CMR imaging to accurately quantify RV volumes. In particular, the identification of RV boundaries near the RV outflow tract may be quite challenging from short-axis slices perpendicular to the long axis of the left ventricle, which is the standard for CMR image acquisition. It is likely that a different acquisition strategy is necessary for accurate RV volume measurements. Indeed, 2 more recent studies using new software designed specifically for volumetric analysis of the right ventricle from RT3DE data sets (Figure 9A) and from a combination of short- and rotated long-axis CMR views found high levels of agreement between the 2 techniques.51,52 Moreover, the RT3DE measurements were both more accurate and more reproducible than several 2DE-based measurements.51Download figureDownload PowerPointFigure 9. Software for volumetric analysis of heart chambers. A, Cast of the RV cavity showing in different colors the inlet septum (green), trabecular septum (pink), and infundibular septum (yellow) of the ventricle. B, 3D surface of the left atrium superimposed on the RT3DE data set.RT3DE Evaluation of Atrial VolumesLeft atrial (LA) enlargement is a marker of both the severity and long-term elevation of LA pressures. It is known to be associated with increased incidence of atrial fibrillation, ischemic stroke, and poor cardiovascular outcomes, including increased risks of overall mortality in patients after MI. When LA size is measured, volume determinations should be preferred over linear dimensions because they allow accurate assessment of the asymmetrical remodeling of the left atrium. Consequently, LA volume calculations from 2D cut planes are recommended as a standard technique53 instead of linear measurements. However, both the proposed area-length technique and the biplane method of disks are dependent on the selection of the location and direction of the LA minor axis, the ability to clearly visualize and accurately trace the LA boundaries, and geometric modeling. With its independence of geometrical assumptions, RT3DE imaging has the potential to provide more accurate LA volume measurements (Figure 9B). However, there is no consensus on the specific methods that should be used for data acquisition and a