Tomographic Three-dimensional Echocardiographic Determination of Chamber Size and Systolic Function in Patients With Left Ventricular Aneurysm
Comparison to Magnetic Resonance Imaging, Cineventriculography, and Two-dimensional Echocardiography
Background Two-dimensional (2D) echocardiographic approaches based on geometric assumptions face the greatest limitations and inaccuracies in patients with left ventricular (LV) aneurysms. Three-dimensional (3D) echocardiographic techniques can potentially overcome these limitations; to date, however, although tested in experimental models of aneurysms, they have not been applied to a series of patients with such distortion. The purpose of this study was therefore to validate the clinical application of tomographic 3D echocardiography (3DE) by the routine transthoracic approach to determine LV chamber size and systolic function without geometric assumptions in patients with LV aneurysms.
Methods and Results In 23 patients with chronic stable LV aneurysms, LV end-systolic and end-diastolic volumes (LVEDV, LVESV) and ejection fraction (LVEF) by tomographic 3DE were compared with results from 3D magnetic resonance tomography (3DMRT) as an independent reference as well as with the conventional techniques of single plane and biplane 2D echocardiography and biplane cineventriculography. Dynamic 3DE image data sets were obtained from a transthoracic apical view with the use of a rotating probe with acquisition gated to control for ECG and respiration (Echoscan, TomTec). Volumes were calculated from the 3D data sets by summating the volumes of multiple parallel disks. 3DE results correlated and agreed well with those by 3DMRT, with better correlation and agreement than provided by other techniques for LVEDV (3DE: r=.97, SEE=14.7 mL, SD of differences from 3DMRT=14.5 mL; other techniques: r=.84 to .93, SEE=30.7 to 41.6 mL [P<.001 versus 3DE by F test], SD of differences=31.5 to 40.7 mL [P<.001 versus 3DE by F test]). The same also pertained to LVESV (3DE: r=.97, SEE=12.4 mL, SD of differences=12.9 mL; other techniques: r=.81 to .90, SEE=24.7 to 37.2 mL [P<.001], SD of differences=27.6 to 36.8 mL [P<.005]) and LVEF (3DE: r=.74, SEE=5.6%, SD of differences=6.7%; other techniques: r=.14 to .59, SEE=9.5% to 10.1% [P<.01], SD of differences=9.5% to 12.6% [P<.05]). Compared with 3DMRT, 3DE was less time consuming and patient discomfort was less.
Conclusions Tomographic 3DE is an accurate noninvasive technique for calculating LV volumes and systolic function in patients with LV aneurysm. Unlike current 2D methods, tomographic 3DE requires no geometric assumptions that limit accuracy.
Determination of global systolic function of asymmetric left ventricles with aneurysm, a common complication of myocardial infarction, poses a methodological problem with the use of conventional 2D imaging techniques.1 Although the analysis of global LV function in this condition is difficult,2 reliable determination of the quantitative parameters of this function is of the utmost clinical importance and is even a precondition for follow-up studies after myocardial infarction to detect LV remodeling3,4 as well as for making the decision for aneurysmectomy,5–7 including predicting the outcome.
To determine LV volumes and LVEF, as the important parameters of global systolic LV function, LVEDV and LVESV must be measured accurately. Until now, asymmetric LV volumes have been approximated by different algorithms calculating symmetric rotation ellipsoids by the use of single or biplane echocardiographic or radiographic imaging.1,2 The following methods have been developed: the ellipsoid method,8 the area-length method,9 and the disk method,10 for single and biplane calculation, respectively. Erbel et al11 correlated all these methods by echocardiographic and radiographic determination of volumes in asymmetric model hearts. Only for the biplane disk method was linearity at ANCOVA found. Nonetheless, by summing up elliptic disks from the two diameters of two orthogonal planes (Fig 1⇓), the biplane disk method can create all approximations close to the true dimensions as well as overestimation, but most often underestimation is observed (Fig 2⇓).11 Erbel et al11 therefore suggested that more than two planes should be used to determine LV dimensions more accurately. Three-dimensional echocardiographic image acquisition, with the use of multiple echocardiographic image planes, has recently been validated for LV volume determination but was limited in the acquisition of equidistant cross sections for using the disk method, as suggested by Weyman,12 as well as in the acquisition of heart rate–triggered dynamic data sets for determination of LV function.13–17
The technical requirements for realizing this idea by means of echocardiographic tomographic imaging are available today. Wollschlaeger et al18 first described a technique for the acquisition of a complete dynamic three-dimensional echocardiographic data set, which allows an unlimited choice of views for all cardiac structures and spaces but also a division in equidistant tomographic slices in a parallel order, as in magnetic resonance or computer tomographic imaging (Fig 1⇑). To test the accuracy of this method for quantifying LV volumes in the presence of aneurysm, our group performed an experimental prestudy with asymmetric model hearts. We compared the volumes calculated from tomographic 3DE to true dimensions obtained by direct measurements.19 This in vitro study revealed good agreement and correlation between 3DE measurements and direct measurements of the true model heart volume without significant underestimation or overestimation but poor agreement and correlation between bp2DE and both 3DE and direct measurements, with strong overestimation of 2D measurements.
After these experiments the purpose of the following study was to evaluate for the first time the accuracy of tomographic 3DE as a clinical application for quantifying global systolic function in a series of patients with LV aneurysm after myocardial infarction. 3DE results were compared with results from 3DMRT as an independent reference that has been validated for determining LV volumes accurately20,21 as well as with the conventional techniques of sp2DE, bp2DE, and bpCVG.
To determine global function of asymmetric left ventricles, LVEDV and LVESV were measured in 23 consecutive patients (age, 63±9.9 years; 14 men, 9 women) with chronic stable LV aneurysms of different size, shape, and localization. In 5 patients the aneurysm was localized in the posteroinferior wall, in 13 patients in the anterior wall, and in 5 patients in the apex of the heart. In 22 patients aneurysm developed after myocardial infarction. The young patient without coronary artery disease developed the large apical aneurysm as a complication of acromegaly. Seventeen patients underwent coronary angiography for clinical reasons and were identified for this study from cineventriculography having a LV aneurysm. Six patients were identified from routine echocardiography. After obtaining informed consent, in the next 2 days all other examinations were performed to establish comparable blood pressure, heart rate, and hemodynamics.
For dynamic 3DE, image data acquisition was performed by a commercially available phased-array sector scanner (model SSH 140 A; Toshiba Corp) and a 3.75-MHz transducer from a transthoracic apical view. 3D image data sets were obtained by rotating the transducer with a special rotation motor device (TomTec) in 2 degree steps and digitizing 90 ultrasound images by the Echoscan system (TomTec) automatically.19 During the examination, patients were placed in a left decubital supine position, and the rotation device was mounted on a stand device, fixing the transducer at the chest wall during rotation. For dynamic 3DE reconstruction, rotation was performed automatically by ECG and respiratory triggering. At every step of rotation 2D images were acquired at every ECG interval of 40 ms and at end expiration, respectively. After digital storage, the 3D data set underwent a postprocessing procedure composed of a gap filling between the acquired images (interpolation) and a conversion of data points (pixels) ordered in a cylindrical coordinate system into a right-angled (Cartesian) coordinate system. For interpolation, average image information from the two images bordering the gap was used as infill. Volumes were calculated from the 3D data sets by summating the volumes of multiple parallel disks (Figs 1⇑, 3⇓, and 4⇓).19 The distance between the tomographic cross-sectional images, identical to the disk thickness, was 2.9 mm.
Tomographic Magnetic Resonance Imaging
Images were obtained with a commercially available 1.5-T superconducting magnetic resonance imaging (MRI) system (Siemens Magnetom SP 63). ECG-gated localizing spin-echo sequences were used to identify the intrinsic long axes of the heart. For all patients, short-axis MRI images were started at the mitral valve plane. Long axes were subdivided in the middle, and the two levels at the same position of each half (1st/1st, 2nd/2nd, etc) were scanned simultaneously with a slice thickness of 8.0 mm. The left ventricle was covered with five or six of those short-axis series.
BpCVG was performed in a digitally equipped catheterization laboratory (HICOR, Siemens). For bpCVG, 35 mL of a nonionic contrast agent (Ultravist, Schering) was injected into the left ventricle with a flow rate of 12 mL/s with the use of a 5F or 7F pigtail catheter. The biplane projections were recorded in 30 degree right anterior oblique (RAO) and 60 degree left anterior oblique (LAO) projections. The afterward image calibration was performed with the use of a metal ball with a diameter of 5.0 cm with identical positions of the x-ray tubes. This was possible with an automatic readjustment to the correct positions using the system memory function.
Monoplane and Biplane Echocardiography
Conventional transthoracic echocardiographic imaging of the left ventricle was performed in the apical RAO equivalent view and for biplane studies the apical RAO and apical two-chamber view.11 Both 2D studies were performed immediately after scanning for the 3D echocardiographic examination with the same phased-array sector scanner (model SSH 140 A; Toshiba Corp) with a 3.75-MHz transducer. Patients remained in the same left decubital supine position as for the 3D examination before. The resolution of the system has been tested previously and demonstrated high accuracy for LV volume and EF determination in patients with LV aneurysm.22
After image acquisition for all of the described methods, LVEDV and LVESV were determined and resulting LVEF was calculated. For both 3DE and 3DMRT, LVEDV and LVESV were calculated with the use of the 3D disk method (Fig 1⇑).19 Biplane volume calculations for bpCVG and echocardiographic image data were performed by the disk method with an external system (ECHO-COM 2.0 for Windows; Individual Software GmbH). Therefore, two biplane images of one-phase end diastole or end systole were displayed side by side on one screen with the use of a special copy function integrated into the echocardiographic unit (Hewlett Packard Sonos 1500). Monoplane echocardiographic volume determination was performed with the use of the area-length method because this is the method most often applied in routine echocardiographic diagnostics.
The methods were compared in accordance with the recommendations of Bland and Altmann.23 The differences (y-axis) were plotted against mean values between methods (x-axis). The degree of agreement between two methods was determined as mean differences (bias), standard deviation of the differences, limits of agreement (mean±2 SD), standard error of the mean difference, and 95% confidence interval of the mean difference. A one-sample t test at the 5% significance level was used to analyze whether the resulting difference from zero, as an underestimation or overestimation by 3DE measurements, was significant. A two-tailed F test at the 5% significance level was used to analyze equality of the standard deviation of differences between 3DE and 3DMRT to the standard deviation of the differences between 3DMRT and the other techniques. Because the repeatability of 3DE measurements is relevant for the amount of agreement, the coefficient of repeatability was determined for two 3DE measurements performed by one observer. In the same way, interobserver variability was tested in a comparison of measurements performed by two observers experienced in that field. Mean differences for repeatability and interobserver variability testing were set at zero because the same method was used.
In a second analysis, linear regression was calculated, specifying the correlation coefficient (r), intercept (a), slope (b), standard error of estimation (SEE), and 95% confidence interval of the regression line. As a reference technique for statistical analysis, independent variables (x-axis) were measured by 3DMRT and dependent variables (y-axis) by 3DE and 2D methods. A two-tailed F test at the 5% significance level was used to analyze the equality of SEEs between the correlation of 3DE to 3DMRT and the correlation of 3DMRT to the other techniques.
Manual border detection of the entire LV endocardial surface was possible in the 18 3D data sets with posteroinferior or anterior wall aneurysms except for one patient. In the remaining five 3D data sets with apical aneurysms, part of the endocardial surface was detected incompletely because of the cone-shaped reconstruction obtained from an apical rotation. Nevertheless, endocardial surface detection was satisfactory enough to include these five patients. Manual border detection in this study was easier and more satisfactory in tomographic 3DE than in 3DMRT. Especially in the aneurysm region, endocardial border delimitation was sometimes difficult in MRI. The average number of cross sections analyzed to calculate LVEDV by 3DE was 31.4±6.3 and for LVESV 29.3±7.2. The average number of cross sections analyzed to calculate volumes by 3DMRT was 11.2±2.1 and 10.1±2.2, respectively. The acquisition time for a 3DE data set was 2.3±1.0 minutes and was dependent on the subject’s heart rate, cardiac rhythm, and respiration. After acquisition of the 3DE data set, the average time for postprocessing and measurement was 34±8.2 minutes. The total image acquisition time for 3DMRT ranged from 40 to 60 minutes and was determined by the patient’s heart rate, cardiac rhythm, and the number of series required. After acquisition of the 3DMRT data set, the average time for postprocessing and measurement was 28±8.4 minutes. No 3DMRT examination had to be broken off because of the development of claustrophobia. The individual results of the 23 patients are shown in the Table 1⇓⇓.
Three-dimensional Echocardiography Versus Three-dimensional Magnetic Resonance Imaging
The results for LVEDV and LVESV followed a normal distribution in the 3DE measurements but not in the 3DMRT measurements; the results for LVEF followed a normal distribution in both the 3DE and 3DMRT measurements. For LVEDV, LVESV, and LVEF, values obtained by 3DE ranged from 78 to 297 mL (mean, 173±62 mL), 53 to 252 mL (mean, 117±53 mL), and 19% to 50% (mean, 33±8%) and ranged from 82 to 300 mL (mean, 183 mL), 56 to 276 mL (mean, 121 mL), and 19% to 54% (mean, 36±10%) for 3DMRT. The differences between the two methods ranged for LVEDV, LVESV, and LVEF from −47 to 11 mL (mean, −10.7±14.5 mL), −41 to 10 mL (mean, −3.4±12.9 mL), and −20.7% to 9.7% (mean, −2.5±6.7%) with SEE of ±3.0 mL, ±2.7 mL, and ±1.4% and 95% confidence interval between −16.6 and −4,7 mL, −8.3 and 1.9 mL, and −5.2% and 0.3%, respectively. Limits of agreement amounted to −39.7 and 18.4 mL, −29.1 and 22.4 mL, and −15.8% and 10.9%, respectively. With P=.002, P=.22, and P=.091, a significant difference of the mean difference from 0 was found for LVEDV as an underestimation of 3DE compared with 3DMRT measurements; however, no significant difference of the mean difference from 0 was found for LVESV and LVEF.
Linear regression analysis was used to correlate LVEDV, LVESV, and LVEF measurements by the two methods. The correlations were described by y=0.97x−5.6 mL, r=.97, SEE=±14.7 mL; y=0.92x+6.2 mL, r=.97, SEE=±12.4 mL; and y=0.61x+11.7%, r=.74, SEE=±5.6%.
The plot in Fig 5a⇓ depicts the finding that the mean difference of LVEDV measurements is near zero, indicating that there was no relevant systematic overestimation or underestimation. Nevertheless, statistical analysis detected a statistically significant underestimation by 3DE compared with 3DMRT. However, the limits of agreement were small, indicating good agreement between the two methods. For LVEF, Fig 6a⇓ shows the mean difference to be near zero with small limits of agreement. Fig 7a⇓ documents the finding of a good correlation between the two methods for LVEDV, which might underline the good agreement between the two methods. For LVEF, Fig 8a⇓ documents low correlation but with acceptable limits of the 95% confidence interval and an acceptably small SEE.
Biplane Cineventriculography and Biplane and Monoplane Echocardiography Versus Magnetic Resonance Imaging
On 2D measurements, only the results of LVEF by all methods and of LVESV by bpCVG were normally distributed. On measurement of LVEDV values for bpCVG, bp2DE and sp2DE ranged from 105 to 429 mL (mean, 223 mL), 95 to 446 mL (mean, 175 mL), and 88 to 397 mL (mean, 162 mL). The mean differences compared with 3DMRT were 49.0±31.5 mL, −8.0±40.7 mL, and −20.9±37.6 mL, with limits of agreement between −14.0 and 111.9 mL, −89.4 and 73.4 mL, and −96.2 and 54.4 mL and 95% confidence intervals between 33.5 and 64.4 mL, −24.6 and 8.7 mL, and −36.3 and −5.5 mL, respectively. SEE was ±7.9 mL, ±8.5 mL, and ±7.9 mL. With P<.001, P=.007, and P=.014 by t test, there was a statistically significant overestimation for LVEDV measured by bpCVG and a statistically significant underestimation by bp2DE and sp2DE. With P<.001, P<.001, and P<.001 by F test, the standard deviations of the differences were significant larger compared with the standard deviation of the differences between 3DE to 3DMRT.
On measurement of LVESV values for bpCVG, bp2DE and sp2DE ranged from 72 to 262 mL (mean, 141±54 mL), 52 to 291 mL (mean, 110 mL), and 42 to 323 mL (mean, 105 mL). The mean differences compared with 3DMRT were 25.3±24.4 mL, −11.2±27.6 mL, and −15.4±36.8 mL, with limits of agreement between −23.4 and 74.0 mL, −66.4 and 44.0 mL, and −89.0 and 58.2 mL and 95% confidence intervals between 13.3 and 37.2 mL, −22.5 and 0.1 mL, and −30.4 and 0.4 mL, respectively. SEE was ±5.1 mL, ±5.8 mL, and ±7.7 mL. With P<.001, P=.064, and P=.057 by t test, there was a statistically significant overestimation for LVEDV measured by bpCVG but no statistically significant overestimation or underestimation for bp2DE and sp2DE measurements compared with results from 3DMRT. With P<.005, P<.001, and P<.001 by F test, the standard deviations of the differences were statistically significant larger compared with the standard deviation of the differences between 3DE to 3DMRT.
On LVEF measurement we found ranges of 9% to 52% (mean, 37±12%), 18% to 60% (mean, 39±9%), and 18% to 57% (mean, 38±11%), respectively. The mean differences compared with results from 3DMRT were 1.5±9.5%, 2.7±12.6%, and 1.8±11.6%, with limits of agreement between −17.5% and 20.4%, −22.4% and 27.9%, and −21.3% and 24.9% and 95% confidence intervals between −3.2% and 6.1%, −2.4% and 7.9%, and −2.9% and 6.5%, respectively. SEE was ±2.4%, ±2.6%, and ±2.4%. With P=.546, P=.303, and P=.466 by t test, no statistically significant overestimation or underestimation was found. With P<.05, P<.001, and P<.005 by F test, the standard deviations of the differences were statistically significant larger compared with the standard deviation of the differences between 3DE to 3DMRT. In bpCVG measurements there was, however, a clear outlier, which was probably due to a very uncommon ventricular shape and was omitted for statistical analysis as a statistically accepted procedure.23
Compared with 3DMRT, the correlation for the LVEDV was described for bpCVG by y=1.17x+19.5 mL, r=.93, and SEE=±30.7 mL; for bp2DE by y=1.00x−8.8 mL, r=.84, and SEE=±41.6 mL; and for sp2DE by y=0.95x−11.3 mL, r=.84, and SEE=±38.4 mL. With P<.001, P<.001, and P<.001 by F test, SEEs were statistically significantly larger compared with the correlation of 3DE to 3DMRT. For LVESV, the correlation was described for bpCVG by y=0.91x+35.7 mL, r=.90, and SEE=±24.7 mL; for bp2DE by y=0.88x+3.5 mL, r=.88, and SEE=±27.4 mL; and for sp2DE by y=0.90x−3.2 mL, r=.81, and SEE=±37.2 mL. With P<.001, P<.001, and P<.001 by F test, SEEs were statistically significantly larger compared with the correlation of 3DE to 3DMRT. For LVEF, the correlation was described for bpCVG by y=0.83x+7.5%, r=.59, and SEE=±9.7%; for bp2DE by y=0.13x+34.0%, r=.14, and SEE=±9.5%; and for sp2DE by y=0.39x+23.7%, r=.36, and SEE=±10.1%. With P<.005, P<.01, and P<.005 by F test, SEEs were statistically significant larger compared with the correlation of 3DE to 3DMRT.
Fig 5b⇑ and d document that neither the mean differences between bpCVG and 3DMRT nor sp2DE and 3DMRT are close enough to zero, indicating a remarkable systematic overestimation for LVEDV measurements by bpCVG and a slight underestimation of sp2DE measurements. Fig 5c⇑ shows that there was no relevant overestimation or underestimation for bp2DE measurements. Only for the 3DE measurements of LVEDV were no outliers found. The bpCVG outlier of a remarkable overestimation of that technique, which was omitted from this analysis, was obtained from the acromegaly patient having an extreme asymmetric left ventricle with a very large apical aneurysm but a normal, tall ventricular base. Thus by bpCVG using the 60 degree LAO projection, the large aneurysm was prominent at the second plane for biplane calculation and its large protrusion was calculated over the whole ventricular shape, but not only for the apex. We also found overestimation for the sp2DE and bp2DE measurements in the same patient, but to a significantly smaller extent, so we decided not to exclude the 2DE outliers. However, it must be taken into account that inclusion of the cineventriculographic outlier would yield much greater limits of agreement and the exclusion of the 2DE outliers much smaller limits of agreement. Figs 6b⇑, 6c⇑, and 6d⇑ document the fact that for LVEF the mean differences between bpCVG, bp2DE and sp2DE compared with 3DMRT are close enough to zero, indicating that there was no relevant overestimation or underestimation. In all three comparisons for LVEDV (Figs 5b⇑, 5c⇑, and 5d⇑) and LVEF (Figs 6b⇑, 6c⇑, and 6d⇑), however, the limits of agreement were wide, indicating unacceptable agreement between the three methods compared with 3DMRT. Fig 7b⇑ documents for bpCVG an acceptable correlation coefficient of r=.93, but the overestimation and an SEE of ±30.7 mL indicate poor correlation compared with 3DMRT. In comparison to Fig 7b⇑, Figs 7c⇑ and 7d show slightly better correlation for the 2DE methods than for bpCVG, but nevertheless the limits of the 95% confidence interval are wide and SEEs are large, also indicating poor correlation. Figs 8b⇑, 8c⇑, and 8d⇑ show the correlations for LVEF, which are poorer for all three 2D methods as approved to 3DMRT.
Accuracy of Repeat Measurements and Interobserver Comparison
For repeat measurements of LVEDV, LVESV, and LVEF by 3DE, we found a mean difference of −0.3±6.2 mL, −0.4±3.9 mL, and 0.4±2.7%, SEE of ±1.3 mL, ±0.8 mL, and ±0.6%, limits of agreement of ±12.4 mL, ±7.8 mL, and ±5.4%, and a 95% confidence interval of ±2.5 mL, ±1.6 mL, and ±1.1%, respectively. For the interobserver measurements of LVEDV, LVESV, and LVEF by 3DE, we found a mean difference of 1.3±7.5 mL, 2.0±5.1 mL, and −0.5±4.8%, SEE of ±1.6 mL, ±1.1 mL, and ±1.0%, limits of agreement of ±9.6 mL, ±10.1 mL, and ±9.4%, and a 95% confidence interval of ±3.1 mL, ±2.1 mL, and ±2.0%, respectively.
These results for LVEDV and LVEF are plotted in Figs 9a⇓, 9b⇓, 9c⇓, and 9d⇓ and reveal good reproducibility and a small interobserver variability of 3DE measurements, which indicates furthermore that tomographic 3DE provides reliable and accurate measurements in clinical studies.
In asymmetric left ventricles, especially those with aneurysms caused by myocardial infarction, accurate quantification of LV volumes and LVEF is of the utmost clinical importance for determination of global LV function. Although the first attempts to use M-mode echocardiography in volume determination in patients with LV aneurysms produced large discrepancies compared with cineangiography,2 improved correlations were later obtained between 2DE and single-plane24,25 and biplane cineventriculography10,26 and radionuclide methods.27 Most of the studies revealed statistically significant systematic underestimation of LV volumes by echocardiography,10,24,25,28 even if biplane methods were used (Fig 1⇑).10,28 Erbel et al11 compared the accuracy of different single-plane and biplane two-dimensional echocardiographic methods and algorithms for volume determination of asymmetric model hearts and found the best correlation with true model heart dimensions with the biplane disk method. They described the limitations of two-dimensional quantification of asymmetric volumes, even by biplane methods, and recommended the development of multiplane quantification. In comparison to echocardiographic findings for biplane cineventriculography, overestimation was documented in previous studies.9,11,29
On the basis of the experience obtained in the in vitro study,19 in this in vivo study tomographic determination of global systolic LV function by 3DE was compared with 3DMRT, sp2DE and bp2DE, and bpCVG in 23 patients with chronic LV aneurysm. We found good agreement only between tomographic 3DE and the 3DMRT measurements of LV volumes and LVEF, and similar to the in vitro results, tomographic 3DE measurements were highly reproducible combined with low interobserver variability. In contrast to previous 2DE studies10,24,25 we found a far smaller but nevertheless significant underestimation for the determination of LVEDV by 3DE but no significant underestimation or overestimation of LVESV and LVEF. For sp2DE, bp2DE, and bpCVG, agreement and correlation with 3DMRT of both LV volumes and LVEF were poor, with significant underestimation of sp2DE and bp2DE measurements of LVEDV and remarkable overestimation of LVEDV and LVESV measured by bpCVG. However, as a more relevant finding than systematic errors for all measurements−LV volumes and LVEF−the standard deviations of the differences between 3DMRT and either bpCVG, bp2DE, or sp2DE were statistically significantly larger compared with the standard deviation of the differences between tomographic 3DE to 3DMRT.
The rotational scanning approach used in this study revealed 3D data sets with high resolution. Compared with 2DE imaging, echo dropouts were less relevant in the 3D reconstruction because of the compensatory effect of information from adjacent wall portions and interpolation. Limitations of the rotational approach for clinical studies include the apical position of the transducer, which tends to be superior to the anatomic apex of the left ventricle,30 and the cone-shaped reconstruction, both potentially limiting the detection of apical aneurysms in some cases, leading to underestimation of LV volumes. The volume underestimation observed for the bp2DE measurements10,28 is in addition caused by the aneurysm shapes deforming the calculated ellipse (Fig 2⇑). Similarly, this relationship leads to the underestimation found for sp2DE volume measurements.
All in all, there is little information in the literature on determination of systolic function with the use of imaging techniques in patients with LV aneurysm, except for use of the cineventriculographic technique as the standard technique over a long period of time. Although previous studies comparing 2D and 3D imaging techniques with cineventriculography have shown good correlation and agreement for volume measurements in normal-shaped left ventricles,30 studies comparing 2D and 3D imaging techniques with cineventriculography have documented the limitation of even bpCVG for determining asymmetric ventricular volumes11,31 which is confirmed by our study.
Comparison to Previous Studies
The tomographic 3D echocardiographic technique used here provides a highly accurate estimate of LV volumes without geometric assumptions with the use of complete 3DE image information on the heart with the ability to calculate 3DE image reconstructions with a realistic gray-scale surface texture. Using this technique, in both normal and pathologically shaped left ventricles, Kupferwasser et al31 found an improved correlation between 3DE and bpCVG measurements of LVEDV and LVESV (y=0.95x−5.9, r=.99; y=0.93x+1.9, r=.99, respectively) as compared with bp2DE and bpCVG measurements (y=0.84x−0.35, r=.98; y=0.86x−1.64, r=.97, respectively). In normal-shaped left ventricles, Nosir et al32 documented excellent correlation of 3DE with radionuclide angiography for the calculation of LVEF (y=0.9x+3.7, r=.99), with very small limits of agreement (−0.39% and 0.32%, bias=−0.035%).
Several quantitative evaluations of LV volumes by three-dimensional echocardiography have been performed by two different techniques, but both depend on a geometric assumption, although a 3D data set is acquired from multiple image planes. The first technique is polyhedral surface reconstruction to estimate ventricular surface and a large number of tetrahedrons reconstructed between image planes for volume computation.14,16 Sapin et al16 also documented accurate volume measurements in asymmetric left ventricles for this technique by using excised porcine hearts. This group also showed better correlation of this 3D technique with single-plane cineventriculography for LV volume estimation compared with bp2DE in patients with normal-shaped left ventricles. In their study they found equivalent measurements for bias and standard deviation of LVEF for 3DE (bias=6.6±9.8%) and bp2DE (bias=7.5±10.7%) compared with cineventriculography, which is different from the findings in the study presented here on patients with LV aneurysm.33 Applying this technique, systolic LV function was also evaluated by Gopal et al34 in unselected patients with suspected heart disease. The results for LVEF in that study comparing 3DE with equilibrium radionuclide angiography are close to the results in our own study comparing tomographic 3DE with 3DMRT. Recently, in young patients with pulmonary hypertension and compressed left ventricles, Apfel et al35 found results similar to those in our study with a slight underestimation of LVEDV and LVESV by 3DE (r=.94 and 0.87, bias=−6.9±6.9 mL and −16±11.2 mL, respectively) compared with 3DMRT, also with a poorer correlation for 2DE. LVEF by 3DE also had closer agreement with 3DMRT (bias=1.1±7.7%) than 2DE (bias=4.4±13.9%). As in our study, the bias results for measurements of LVEF were close to zero for 3DE as well as for 2DE, but the standard deviation as the more critical parameter was significantly larger for 2DE. The second technique combines the digitized video images with their spatial locations obtained by the spark gap triangulation system.15,17 Jiang et al36 validated the spark-gap technique for the determination of LV volumes in the presence of aneurysm and found good agreement of calculated in vitro volumes with actual values. In the same study in an in vivo animal model, Jiang et al showed good correlation of this 3D technique with actual volume values (LVEDV: r=.99, SEE=4.3 cm3; LVESV: r=.99, SEE=3.5 cm3) and LVEF (r=.99, SEE=2%) of ventricles with aneurysm. A slight underestimation was found for LVEDV, with a correlation of y=1.04x−4.2, similar to the study presented (y=0.97x−5.6).36
Limitations of the Study
No relevant differences in the image quality of the cross-sectional images between tomographic 3DE and 3DMRT were found in our study, providing manual planimetry with the same accuracy. However, 3DMRT imaging in this special study population was limited by reduced blood pool detection in the aneurysms caused by blood stasis and therefore limited delineation of the endocardial border in this region.
Even though the acquisition of a 3DE data set provides accurate representation of the LV cavity shape, especially in asymmetric ventricles, there are few limitations that lead to minor underestimation. For the tomographic 3DE technique used in our study, the reason for this underestimation was mainly the cone-shaped data set with limited detection of apical aneurysms. For 3DE techniques using multiple image planes of the left ventricle from different views with the creation of wire-frame models of the ventricle,15,33 the reason for the underestimation is mainly the need for volume computation using geometric assumptions with the tendency to round off sharply protruding regions, as discussed by Pearlman37 previously as a serious limitation.
Manual endocardial border tracing in both tomographic techniques was a serious limitation because of the time factor. Integration of an automated border detection technique is a necessary prerequisite for the application of volume calculations by tomographic 3DE in clinical routine diagnoses. Another limitation was the image frequency of 25 images/s, as provided by the 3D system used in this study. This system does not allow selection of the image representing the exact end-diastolic and end-systolic points of time. Thus both 2D echocardiographic systems and 3D reconstruction systems need to be able to process higher image frequencies.
The tomographic 3DE technique used in this study for the first time provides noninvasive clinical information on patients with LV aneurysm before or after aneurysmectomy: accurate quantification of LV volumes and systolic function as well as dynamic 3D visualization of the real anatomy of the ventricle and the aneurysm. Previous studies have described the relationship of residual myocardium to the aneurysm area in standard 2DE projections, but accurate determination of heart size, ventricular function, and aneurysm size has not been available.5 In comparison, with 3DE imaging it should be possible also to determine size, shape, localization, and dynamic characteristics of LV aneurysms, as shown in the first preliminary results in human subjects.38 This would provide preoperative information for the surgeon about patch size and make it possible to calculate the resulting ventricular size, shape, and function after aneurysmectomy. Furthermore, with the use of 2D techniques, the number of prognostic parameters before aneurysmectomy is too small for reliable preoperative patient selection, on the basis of noninvasive data.6 After the clinical validation presented, the next step in clinical studies is to determine whether tomographic 3DE will improve patient selection and decision making before aneurysmectomy, intraoperative guiding, perioperative monitoring, and postoperative control, as well as the prediction of outcome. To monitor a patient for these criteria would require many repeat examinations. Therefore it would be more ideal to use tomographic 3DE than 3DMRT because it is less time consuming, less costly, and patient discomfort is less.
The purpose of this study was to validate the clinical application of tomographic 3DE by the routine transthoracic approach to determine LV volumes and systolic function without geometric assumptions in patients with LV aneurysms. Results from tomographic 3DE correlated and agreed well with those by 3DMRT, with better correlation and agreement than provided by current 2D techniques. Unlike current 2D techniques, 3DE required no geometric assumptions that limit accuracy. Tomographic 3DE was shown to be an accurate technique that can be used in clinical routine because it is noninvasive and less time consuming than other tomographic techniques. This is the basis for further clinical studies and will probably point to new procedures for preoperative selection and perioperative monitoring of patients for aneurysmectomy as for follow-up studies after myocardial infarction to detect LV remodeling. Moreover, future developments to improve the clinical applicability of 3DE volume quantification to routine practice must include automated contour detection for faster volume detection.
Selected Abbreviations and Acronyms
|2D, 3D||=||two-dimensional, three-dimensional|
|2DE, 3DE||=||2D, 3D echocardiography|
|3DMRT||=||3D magnetic resonance tomography|
|LVEDV||=||LV end-diastolic volume|
|LVEF||=||LV ejection fraction|
|LVESV||=||LV end-systolic volume|
|sp 2DE||=||single plane 2DE|
This study was supported in part by a grant from the “Ernst und Bertha Grimmke” Donation, Duesseldorf, Germany.
- Received May 30, 1997.
- Revision received August 19, 1997.
- Accepted September 7, 1997.
- Copyright © 1997 by American Heart Association
Picard MH, Wilkins GT, Ray PA, Weyman AE. Natural history of left ventricular size and function after acute myocardial function. Circulation. 1990;82:484–494.
Cohen M, Packer M, Gorlin R. Indications for left ventricular aneurysmectomy. Circulation. 1983;67:717–722.
Greene DG, Carlisle R, Grant C, Bunell IL. Estimation of left ventricular volume by one-plane cineangiography. Circulation. 1967;35:61–69.
Erbel R, Krebs W, Henn G, Schweizer, Richter HA, Meyer J, Effert S. Comparison of single-plane and biplane volume determination by two-dimensional echocardiography. Eur Heart J. 1978;3:469–480.
Weyman AE. Cross-Sectional Echocardiography. Philadelphia, Pa: Lea & Febiger; 1982.
Handschumacher MD, Lethor JP, Siu SC, Mele D, Rivera JM, Picard MH, Weyman AE, Levine RA. A new integrated system for three-dimensional echocardiographic reconstruction: development and validation for ventricular volume with application in human subjects. J Am Coll Cardiol. 1993;21:743–753.
Siu SC, Rivera M, Guererro JL, Handschumacher MD, Lethor JP, Weyman AE, Levine RA, Picard MH. Three-dimensional echocardiography: in vivo validation for left ventricular volume and function. Circulation. 1993;88:1715–1723.
Wollschlaeger H, Zeiher AM, Klein HP, Kasper W, Geibel A, Wollschlaeger S. Transesophageal echo computer tomography: a new method for dynamic 3-D imaging of the heart. Circulation. 1989;80(suppl II):II-569. Abstract.
Carr KW, Engler RL, Forsythe JR, Johnson AD, Gosink B. Measurement of left ventricular ejection fraction by mechanical cross-sectional echocardiography. Circulation. 1979;59:1196–1206.
Bommer W, Chum T, Kwan L, Neumann A, Mason DT. Biplane apex echocardiography versus biplane cineangiocardiography in the assessment of left ventricular volume and function: validation by direct measurements. Am J Cardiol. 1980;45:471. Abstract.
Starling M, Crawford M, Levi B, Sorenson S, Richards K, O’Rourke R. Comparative efficacy of cross-sectional echocardiography and gated equilibrium radionuclide angiography for assessing left ventricular size and performance. Am J Cardiol. 1980;45:403. Abstract.
Schiller NB, Acquatella H, Ports TA, Drew D, Goerke J, Ringertz H, Silverman NH, Brundage B, Botvinick EH, Boswell R, Carlsson E, Parmley WW. Left ventricular volume from paired biplane two-dimensional echocardiography. Circulation. 1979;60:547–555.
Erbel R, Schweizer P, Lambertz H, Henn G, Meyer J, Krebs W, Effert S. Echoventriculography: a simultaneous analysis of two-dimensional echocardiography and cineventriculography. Circulation. 1983;67:205–215.
Kupferwasser I, Mohr-Kahaly S, Wittlich N, Erbel R, Meyer J. Volumetry in three-dimensional echocardiography using ECHO-CT. Eur Heart J. 1994;15:441. Abstract.
Nosir YF, Fioretti PM, Vletter WB, Boersma E, Salustri A, Postma JT, Reijs AE, Ten Cate FJ, Roelandt JRTC. Accurate measurement of left ventricular ejection fraction by three-dimensional echocardiography: a comparison with radionuclide angiography. Circulation. 1996;94:460–466.
Gopal AS, Shen Z, Sapin PM, Keller AM, Schnellbaecher MJ, Leibowitz DW, Akinboboye OO, Rodney RA, Blood DK, King DL. Assessment of cardiac function by three-dimensional echocardiography compared with conventional noninvasive methods. Circulation. 1995;92:842–853.
Apfel HD, Shen Z, Gopal AS, Vangi V, Solowiejczyk D, Altmann K, Barst RJ, Boxt LM, Allan LD, King DL. Quantitative three-dimensional echocardiography in patients with pulmonary hypertension and compressed left ventricles: comparison with cross-sectional echocardiography and magnetic resonance imaging. Heart. 1996;76:350–354.
Jiang L, De Prada JAV, Handschumacher MD, Vuille C, Guererro L, Picard MH, Joziatis T, Fallon JT, Weyman AE, Levine RA. Quantitative three-dimensional reconstruction of aneurysmal left ventricles: in vivo and in vitro validation. Circulation. 1995;91:222–230.
Buck T, Hunold P, Tkalec W, Nesser HJ, Erbel R. Determination of left ventricular aneurysm surface area by dynamic three-dimensional echocardiography. J Am Coll Cardiol. 1997;29:119A. Abstract.