Three-Dimensional Echocardiographic Estimation of Infarct Mass Based on Quantification of Dysfunctional Left Ventricular Mass
Background Two-dimensional echocardiography is useful for estimating the extent of infarct-related wall motion abnormalities. Such estimation, however, is based on a few selected views and extrapolated for the whole left ventricle (LV). This approach does not provide us with the actual amount of dysfunctional myocardium. Volume-rendered three-dimensional echocardiography (3DE) might overcome these limitations. In this study we explored (1) how well volume-rendered 3DE delineates regional dysfunction of the infarcted LV and (2) how well dysfunctional myocardial mass quantified by 3DE reflects the actual anatomic infarct mass.
Methods and Results 3DE was performed before and 3 hours after coronary occlusion in 16 dogs. With the LV viewed in equidistant short-axis slices, the region of dysfunction was demarcated, and the dysfunctional myocardial mass was derived from this. With triphenyltetrazolium chloride staining, anatomic infarct regions were delineated, dissected, and weighed. The anatomic infarct mass was 16.3±7.7 g (mean±SD) (range, 6.4 to 31.4 g); the dysfunctional mass estimated by 3DE was 17.4±9.1 g (range, 5.2 to 39.0 g). The mean difference was 1.0 g. The correlation between dysfunctional mass (y) and infarct mass (x) was y=1.1x−0.6, r=.93 (P<.0001). The percentage of LV involved in infarction was 18.2±5.8% (range, 9.1% to 26.1%); the percentage of LV involved in regional dysfunction was 18.3±6.9% (range, 7.9% to 31.2%). The mean difference was 0.1%. The correlation between percentage of LV involved in infarction (x) and percentage of LV involved in dysfunction (y) was y=1.0x−1.1, r=.92 (P<.0001).
Conclusions Volume-rendered 3DE crisply displays regional dysfunction of infarcted LV. 3DE-measured dysfunctional mass accurately reflects the anatomic infarct mass.
Accurate estimation of myocardial infarct size is known to have prognostic and therapeutic implications.1 2 2DE has become the most commonly used technique to identify regional myocardial dysfunction caused by infarction and to estimate the extent of dysfunctional LV myocardium. 2DE assessment of dysfunctional myocardium has been shown to correlate with infarct size in both experimental and clinical studies.3 4 5 6 7 However, most such studies used a few selected 2D views in short-axis or apical orientations, from which the extent of the wall motion abnormality was computed. A percentage of dysfunctional myocardium was often either derived for a given 2D slice or extrapolated for the whole ventricle.8 9 10 11 Effort has not been directed toward evaluating how well the extent of regional dysfunctional myocardium reflects the actual anatomic mass of the infarcted myocardium. One of the major reasons for this is the inability of 2DE, hampered by the limited number of imaging planes available to examine a 3D organ, to accurately measure LV mass and, in particular, the mass of a given myocardial region. This difficulty is compounded when the infarcted region undergoes expansion, which leads to regional geometric distortion of the LV. Such distortion could lead to errors when one attempts to extrapolate functional infarct size from limited 2D views. If the whole ventricle could be interrogated with more imaging samples, the resulting 3D data could aid in more accurate quantification of dysfunctional myocardium. Although attempts have been made to use 3D reconstruction techniques to examine ischemic myocardial dysfunction, they have not been applied widely because they were rather laborious and did not allow visualization of the LV myocardium. Volume-rendered 3DE has recently been shown to be clinically feasible.12 13 14 This technique is capable of reproducing dynamic cardiac anatomy in all its dimensions and also yielding quantitative data. Preliminary studies have suggested that it might be possible to measure myocardial mass.15 However, the utility of 3DE in qualitative depiction of wall motion abnormalities and in quantitative estimation of the extent of dysfunctional myocardium is not known. This study was designed to address the following questions: (1) how well volume-rendered 3DE demonstrates regional myocardial dysfunction in 3D display projections and (2) how well 3DE quantification of dysfunctional LV mass reflects actual anatomic infarct mass.
We used 3DE in an open-chest canine model of acute myocardial infarction. With a rotational mode of data acquisition, 3D data sets were collected, 3D reconstructions were performed, and display projections were derived from various orientations to identify regional dysfunction. The mass of the whole LV and that of the dysfunctional region were quantified from the 3D data set. TTC staining was used to delineate and quantify infarcted regions in the autopsied hearts. The mass of the whole LV and that of the infarcted region were determined. The echocardiographic data were compared with the actual anatomic data.
Sixteen mongrel dogs (22±4 kg in weight) were sedated with acepromazine (20 to 30 mg IM), anesthetized with sodium pentobarbital (25 mg/kg body wt IV), intubated, and ventilated with room air by a volume cycle respirator. Lead II of the ECG was monitored throughout the experiment. A femoral artery and vein were instrumented with fluid-filled catheters for monitoring of arterial pressure and fluid and drug administration. The chest was opened by midline sternotomy. A pericardial cradle was created, and the heart was exposed. One of the main coronary arterial branches or its major secondary branches (left anterior descending coronary artery in 10 dogs and posterior descending coronary artery in 6) was isolated and occluded with a silk snare. Lidocaine (1 mg/kg bolus before coronary occlusion and 0.5 mg/min continuous infusion thereafter) was given intravenously to prevent ventricular fibrillation. Three hours after coronary occlusion, a water bath was arranged above the pericardial cradle for ultrasound transmission without affecting the hemodynamics of the heart. Data acquisition for 3DE was then performed. At the end of the experiment, the dog was euthanized with an intravenous injection of 10 mL potassium chloride (10%), and TTC staining was performed for measurement of the anatomic mass of infarcted myocardium.
Measurement of Total LV Mass and Infarct Mass
A solution of 1% TTC was prepared at a temperature of ≈37°C, and 250 mL was perfused into the root of the aorta at a pressure of 120 mm Hg immediately after the dog was killed, while the ascending aorta was clamped. The heart was explanted, and the LV was isolated, weighed, and cut into 6 to 8 parallel transverse slices; each slice was ≈1 cm thick. The infarct region was defined visually on the basis of its pale appearance in contrast to the noninfarct area stained red by TTC (Fig 1⇓). After the infarct region was delineated by examination of both sides of each slice, the infarct zones were carefully dissected out in each slice and weighed. From the total LV mass and the infarct mass data, the percentage of LV involved in infarction was calculated.
Data Acquisition for 3DE
A commercially available ultrasound unit (Sonos 2500, Hewlett-Packard) was used for 2D image acquisition. This instrument was interfaced with a commercially available 3D image processing system (EchoScan, version 3.0, TomTec Imaging Systems) for on-line data acquisition and storage and off-line data processing, 3D reconstruction, display, and quantification. A carriage device with a rotational motor was mounted onto a 2.5/5 MHz imaging transducer, which was then positioned to image the heart through the water bath from an anterior epicardial orientation. The rotation of the transducer and thus that of the imaging plane were controlled by the 3D system in a predefined manner. 2D imaging of the LV was initiated in a long-axis orientation, and images were obtained at every 3° from 0° through 180°. In 8 dogs, additional data acquisitions were performed at intervals of 1°, 2°, and 5°, respectively. ECG and respiratory gating were used for spatial and temporal registration of the images. The acquired data were calibrated, reformatted, and stored in the computer and transferred onto optical laser disks for off-line processing and analysis.
Echocardiographic Data Processing, 3D Reconstruction, Display, and Quantification
The acquired ultrasound data were postprocessed and interpolated into a voxel-rendered 3D data set. 3D images of the LV before and after coronary occlusion were reconstructed with different cutting planes and projections. The feasibility and ease of displaying and identifying regional LV dysfunction from longitudinal, sagittal, and coronary sections were assessed.
3DE quantification of dysfunctional mass was performed by a blinded observer. The 3D data set of the LV was electronically segmented into 12 to 15 equidistant slices (5.6 to 5.8 mm thick) in short-axis orientation for computation of total LV mass and for calculation of dysfunctional myocardial mass (Fig 2A⇓). For determination of total LV mass, each short-axis slice was reviewed in real time and frame by frame. The LV epicardial and endocardial borders were traced at end diastole with a trackball and digitizing system integrated into the 3D processing computer. The cross-sectional area of the LV myocardium obtained between the epicardial and endocardial contours was given a computer-derived “label” (Fig 2B⇓). By integration of the slice thickness with the slice area, the volume of each slice and the total volume of all myocardial slices were computed in an automated manner by the following quantification algorithm: volume (mL)=Σ (area of each slice [mm2]×slice thickness [mm]). Myocardial volume multiplied by assumed specific gravity (1.04 g/mL) provided myocardial mass (g). For measurement of the mass of the dysfunctional myocardium, each paraplane short-axis slice was viewed in a dynamic mode. Discrete akinetic or dyskinetic segments were identified, a contour was drawn around them from endocardium to epicardium (Fig 2C⇓), and a label was derived (Fig 2D⇓). The volume and mass of the dysfunctional segments (the labeled regions) were calculated as described above. From these data, the percentage of the total LV myocardium involved in regional myocardial dysfunction was calculated. The 3D data set was processed, and quantitative data were derived weeks apart in a blinded manner for evaluating intraobserver variability. Another investigator analyzed the data for obtaining interobserver variability.
To identify the site of infarction, the transverse cut in which the papillary muscles were largest was examined both in anatomic specimens and in short-axis images from the 3D data set by two blinded observers. A transparent overlay with 16 equally spaced radii was used to divide the specimens and images into 16 segments. The zero point was located at the anterior ventriculoseptal junction, and the center of the grid was placed at the center of the LV cavity. Segments with evidence of infarction by TTC staining and akinesis or dyskinesis by echocardiography were identified and compared. One hundred twelve segments in 7 dogs were analyzed for this purpose.
Echocardiographic and anatomic data are expressed as mean±SD. To detect differences between 3DE measurements with anatomic data, we used Student’s paired t test. Data were compared by simple linear regression, and the mean differences and limits of agreement were analyzed by Bland and Altman’s method. Interobserver and intraobserver variability are expressed as the coefficient of variance. For the above analyses, a value of P<.05 was considered significant. To detect differences between 3DE acquired with different degree increments, ANOVA was used with a Bonferroni-Dunn correction, for which a value of P<.005 was considered significant.
3DE Display of Ischemic Myocardial Dysfunction
This study yielded good-quality 3D reconstructions in all experiments. Myocardial regions that exhibited regional dysfunction could be well identified in dynamic 3D projections. The LV before coronary occlusion showed normal contraction and wall thickening in all regions during systole in all 16 dogs. Regions of dysfunctional myocardium after coronary occlusion displayed various regional wall motion abnormalities in volume-rendered dynamic 3D images (Figs 3⇓ and 4⇓). Dynamic displays also demonstrated regional cavity dilation in all dogs after coronary occlusion. In addition to dynamic displays, extraction of the whole myocardium and dysfunctional regions could be performed in all dogs (Fig 5⇓). Such displays allowed a direct 3D perception of the location and extent of dysfunctional myocardium. The dysfunctional territories exhibited various sizes, shapes, and locations in different dogs, depending on the site of coronary occlusion.
Comparison of 3DE Quantification of Dysfunctional Myocardium With Measurements From Anatomic Heart Specimens
TTC staining demonstrated evidence of infarction in all canine hearts. Total LV mass was 87±21 g (range, 54 to 123 g). The mass of infarcted myocardium was 16.3±7.7 g (range, 6.4 to 31.4 g), and the mass of dysfunctional myocardium determined by 3DE was 17.4±9.1 g (range, 5.2 to 39.0 g) (P=NS) (Fig 6A⇓). The percentage of the LV mass involved in infarction based on TTC staining was 18.2±5.8% (range, 9.1% to 26.1%) and was not significantly different from the percentage of dysfunctional myocardium derived from 3DE (18.3±6.9%; range, 7.9% to 31.2%) (P=NS) (Fig 6B⇓).
The correlation between 3DE (y) and anatomic measurements (x) in the determination of total LV mass was y=0.8x+7.3, r=.96, P<.0001. The mean difference between these two methods was 1.2 g (P=NS). The correlation between dysfunctional mass determined by 3DE (y) and infarct mass derived from TTC staining (x) was y=1.1x−0.6, r=.93, P<.0001. The difference between these two methods was 1.0±3.3 g (P=NS) (Fig 7⇓). The correlation between the percentage of LV involved in dysfunction as determined by 3DE (y) and in anatomic infarction determined by TTC staining (x) was y=1.0x−1.1, r=.92, P<.0001. The difference between these two methods was 0.1±3.2%, P=NS (Fig 8⇓).
When we analyzed the measurements of dysfunctional mass, LV myocardial mass, and percentage of dysfunctional myocardium obtained with different imaging intervals (1°, 2°, 3°, and 5°), we observed that there was no difference in the correlations with anatomic infarct mass. Mean differences were not significant (Table 1⇓). No significant difference was found between 3D measurements of data collected with different degree increments.
In 7 dogs, 112 LV segments (from seven slices, one slice in each dog) were analyzed by both 3D and anatomic methods. Among these segments, 29 showed evidence of infarction by TTC staining. Twenty-eight segments on 2D images of the 3D data set showed regional dysfunction (Table 2⇓). The predictive accuracy for infarct location by 3DE was 90%. The discordance between echocardiographic and anatomic identification of the infarct segments was limited to one adjacent segment in each study, which may have been due to the difference in definition of the zero point.
Intraobserver and Interobserver Variability of Quantitative Analysis of 3D Data
When dysfunctional myocardial mass was determined with 3DE, the intraobserver variability was 2%, and the difference between two measurements was 0.3±0.7 g. The intraobserver variability for determining the percentage of dysfunctional LV myocardium was 2%. The difference between these measurements was 0.3±0.9%, with no statistical significance. The interobserver variability for determining dysfunctional mass was 7.9%. The difference between measurements by two observers was 1.8±1.9 g (P=NS). For quantifying the percentage of LV myocardial mass involved in dysfunction, the interobserver variability was 11%. The difference between two observations was 1.6±3.1% (P=NS).
Our animal study demonstrates that the actual mass of infarcted myocardium can be determined on the basis of in vivo volume-rendered 3DE quantification of the mass of dysfunctional myocardium in the setting of acute coronary occlusion. There was an excellent correlation between the mass of dysfunctional myocardium and pathological infarct mass without systematic overestimation or underestimation. Furthermore, this 3DE method yielded an accurate measure of the percentage of LV myocardium that was dysfunctional and thus the percentage of infarcted myocardium. In contrast to previous echocardiographic studies that derived the percentage of infarcted myocardium by extrapolation from a few 2D slices, 3DE allowed analysis of multiple parallel and equidistant slices and thus more reliable quantification of the mass of dysfunctional myocardium. In addition to quantitative information, this technique also yielded dynamic 3D projections of the LV in numerous cutting planes and aided in visual appraisal of the wall motion abnormalities and infarct-induced alterations in regional shape.
2DE in the Evaluation of the Extent of Myocardial Infarction
If accurately measured, the extent of regional myocardial dysfunction could be used to assess the size of myocardial infarction. Among several imaging techniques used to assess the effects of infarct on regional LV function, 2DE has been used extensively for quantitative assessment of regional wall motion abnormalities, and various methods have been used to obtain a quantitative estimate of regional dysfunction.1 2 3 4 5 6 7 8 9 10 11 In many studies, the LV was imaged in a few 2DE views (usually 3 or 4 short-axis views and 1 or 3 apical views), and the circumferential extent of dysfunctional myocardium was determined on the basis of a quantitative or semiquantitative method. From such measurements, the fraction of LV that was dysfunctional was extrapolated. Such data were then correlated to infarct size determined pathologically. Pandian et al5 showed that the fraction of the LV that was dyskinetic correlated well with the anatomically determined infarct fraction of LV (r=.92 and .94 at 20 minutes and 2 days after coronary occlusion, respectively). Weiss et al9 demonstrated a correlation coefficient of .90 between the circumferential extent of myocardial akinesis and dyskinesis and the circumferential extent of transmural infarction. Guyer et al11 compared the percentage of endocardium with abnormal wall motion with that of the endocardial surface overlying histochemically determined infarction; a correlation of r=.86 was obtained. Although a good correlation between the proportion of dysfunctional LV myocardium and the percentage infarct size was shown in these studies, it has not been possible previously to estimate the actual infarct mass in grams. The disadvantages of these conventional approaches include the following. (1) Only a few 2DE views were used to extrapolate for the whole LV. (2) Internal anatomic landmarks were used for obtaining the short-axis images, and slice distances were impossible to determine. Furthermore, the short-axis images were usually recorded from one acoustic window by tilting of the probe; therefore, the acquired images were not truly parallel. (3) During the cardiac cycle, the heart rotates and moves transversely and longitudinally in the thoracic cavity, complicated by movements caused by respiration and leading to errors in analysis of 2D slices.
Advantages of Volume-Rendered 3DE in Myocardial Infarction
3DE overcomes the drawbacks of 2DE in quantification of infarct-related dysfunctional myocardium in the following aspects. Systematic stepwise data acquisition permits imaging of the whole ventricle. The interpolated 3D data set can then be electronically segmented into equidistant parallel slices, which enables automatic volume computation with a computer algorithm. The extrinsic movement of the heart caused by respiration can be minimized with the application of respiratory gating using thoracic impedance. Geometric assumptions often used in 2DE methods are made unnecessary.
The volume-rendered 3DE approach we used has particular strengths. Studies that used 2D data acquisition guided by position-locator devices did not yield dynamic 3D projections and required extensive border tracing for derivation of quantitative data.16 In an in vitro study in which pins were placed on the myocardium to simulate “infarct areas,” good correspondence was shown between the 3D surface area and the simulated infarct area.17 However, this algorithm has not been validated in vivo for quantification of infarct-related myocardial dysfunction. Furthermore, this study did not provide tissue depiction in 3D projections.
With volume-rendered 3DE, gaps between 2D image slices can be interpolated and pixels turned into voxels while the characteristic appearance of cardiac tissue is retained in gray scale. Dynamic 3D images can be reconstructed without any tracing of the cardiac silhouettes on 2D images. Manual labeling is required only for quantitative data. With the application of various shading techniques such as distance, texture, and gradient of the examined object, the reconstructed 3DE images portray cardiac structures more realistically.18 Furthermore, volume-rendered 3DE has the ability to display cardiac images in a dynamic mode. This allows visual appraisal of global and regional LV function, detection of wall motion abnormalities, and aneurysmal deformations. Another important strength of our approach is the ability to quantify the mass of the whole ventricle and that of the dysfunctional region.
An interesting observation in our 3DE study is the lack of overestimation of infarct size based on the quantification of dysfunctional mass; this is in contrast to previous 2DE observations. This is intriguing to us. Although there is no clear answer to why it is so, we feel that this could be explained by the following. (1) In quantification of regional dysfunction, only discretely akinetic or dyskinetic segments were included. Previous 2DE investigations often included hypokinetic segments in their analysis. (2) The actual mass of dysfunctional region was determined in our study, whereas 2DE studies used extrapolation of percent dysfunction per slice or for the whole ventricle. (3) We used a multitude of equidistant parallel slices for determining the mass of regional dysfunction. In previous studies, only a few short- or long-axis slices were analyzed, and often they were not truly tomographic, parallel, or equidistant because images were often derived by tilting or rotation of the transducer relying on internal landmarks. (4) The infarct was transmural in all our dogs, whereas many past studies included nontransmural infarcts as well. (5) In determining an anatomic infarct after TTC staining, we dissected the infarct regions and weighed them; in most previous studies, the infarct regions were traced, the areas were measured, and infarct size was indirectly extrapolated. We feel that these methodological differences between our study and previous investigations could explain the lack of overestimation of infarct size by our 3DE approach.
One may consider that, for accurate 3DE quantification, data acquisition at the smallest interval (such as at 1°), with more samples, would provide more accurate measurement than those collected with larger intervals. Our data in the last 8 dogs demonstrated that a 3° interval for data acquisition is as good as 1° and 2° intervals for accurate measurement of dysfunctional myocardial mass, even in the setting of small infarcts. Acquisition with 5° intervals also yielded good correlation with anatomic measurements, especially in quantifying LV mass.
Limitations of Our Study
There are some limitations in this study. (1) In our open-chest dog experiments, we used an anterior epicardial window for image acquisition. How well 3DE data collection from the parasternal and apical acoustic windows provides reliable quantitative information cannot be determined from this study. (2) Although an excellent relationship between dysfunctional mass and infarct mass was shown in this study, the effect of reperfusion on such a relationship cannot be derived from this study. This, however, could be addressed separately in a reperfusion model. (3) In our series of dogs, all infarcts were transmural. Our observations on the estimation of infarct mass cannot be directly applied to nontransmural infarcts. This important aspect requires further investigation. (4) Although computation of the mass of labeled dysfunctional regions was done automatically by the image processing unit, the demarcation of regional dysfunction on paraplane 2D slices was performed manually because automated edge-detection software for 3D data analysis was not available. Such software is being developed and could, in the future, make the analysis easier. Despite visual delineation of dysfunctional areas, the method has provided excellent estimates of dysfunctional and infarct masses. (5) This study was not designed to provide a segment-by-segment echocardiographic versus anatomic comparison. To verify that our identification of dysfunctional myocardium was correct, we performed such an analysis in a single slice in which anatomic landmarks were identified most reliably. This could have biased our segmental data to a certain degree.
Demonstration that 3DE can yield quantitative measures of the mass of whole LV myocardium, dysfunctional region, and thus infarct size has important investigative and clinical implications. This method could be used to study the effects of physiological, pharmacological, and therapeutic interventions on infarcted myocardium in a more versatile manner than hitherto feasible. Dynamic volume-rendered 3D display and accurate quantification of global and regional LV function could be of value in patients with myocardial infarction and a variety of other pathophysiological scenarios; this requires clinical investigation.
Selected Abbreviations and Acronyms
|2DE||=||two-dimensional echocardiography, echocardiographic|
|3DE||=||three-dimensional echocardiography, echocardiographic|
|LV||=||left ventricle, left ventricular|
- Received January 9, 1997.
- Revision received March 6, 1997.
- Accepted March 11, 1997.
- Copyright © 1997 by American Heart Association
Kan G, Visser CA, Koolen JJ, Dunning AJ. Short and long term predictive value of admission wall motion score in acute myocardial infarction: a cross-sectional echocardiographic study of 345 patients. Br Heart J. 1986;56:422-429.
Gallagher KP, Kumada T, Koziol JA, McKown MD, Kemper WS, Ross J Jr. Significance of regional wall thickening abnormalities relative to transmural myocardial perfusion in anesthetized dogs. Circulation. 1980;62:1266-1274.
Lieberman AN, Weiss JL, Jugdutt BI, Becker LC, Bulkley BH, Garrison JG, Hutchins GM, Kallman CA, Weisfeldt ML. Two-dimensional echocardiography and infarct size: relationship of regional wall motion and thickening to the extent of myocardial infarction in the dog. Circulation. 1981;63:739-746.
Pandian NG, Koyanagi S, Skorton DJ, Collins SM, Eastham CL, Kieso RA, Marcus ML, Kerber RE. Relations between 2-dimensional echocardiographic wall thickening abnormalities, myocardial infarct size and coronary risk area in normal and hypertrophied myocardium in dogs. Am J Cardiol. 1983;52:1318-1325.
Pandian NG, Skorton DJ, Collins SM, Koyanagi S, Kieso R, Marcus ML, Kerber RE. Myocardial infarct size threshold for two-dimensional echocardiographic detection: sensitivity of systolic wall thickening and endocardial motion abnormalities in small versus large infarcts. Am J Cardiol. 1985;55:551-555.
Otto CM, Stratton JR, Maynard C, Althouse R, Johannessen KA, Kennedy JW. Echocardiographic evaluation of segmental wall motion early and late after thrombolytic therapy in acute myocardial infarction: the western Washington tissue plasminogen activator emergency room trial. Am J Cardiol. 1990;65:132-138.
Parisi AF, Moynihan PF, Folland ED, Feldman CL. Quantitative detection of regional left ventricular contraction abnormalities by two-dimensional echocardiography, II: accuracy in coronary artery disease. Circulation. 1981;63:761-767.
Weiss JL, Bulkley BH, Hutchins GM, Mason SJ. Two-dimensional echocardiographic recognition of myocardial injury in man: comparison with postmortem studies. Circulation. 1981;63:401-408.
Force T, Kemper A, Perkins L, Gilfoil M, Cohen C, Parisi AF. Overestimation of infarct size by quantitative two-dimensional echocardiography: the rule of tethering and of analytic procedures. Circulation. 1986;73:1360-1368.
Guyer DE, Foale RA, Gillam LD, Wilkins GT, Guerrero JL, Weyman AE. An echocardiographic technique for quantifying and displaying the extent of regional left ventricular dyssynergy. J Am Cardiol. 1986;8:830-835.
Pandian NG, Nanda NC, Schwartz SL, Fan P, Cao QL, Sanyal R, Hsu TL, Mumm B, Wollschlager H, Weintraub A. Three-dimensional and four-dimensional transesophageal echocardiographic imaging of the heart and aorta in humans using a computed tomographic imaging probe. Echocardiography. 1992;9:677-687.
Delabays A, Pandian NG, Cao QL, Sugeng L, Marx G, Ludomirski A, Schwartz SL. Transthoracic real-time three-dimensional echocardiography using a fan-like scanning approach for data acquisition: methods, strength, problems, and initial clinical experience. Echocardiography. 1995;12:49-59.
Delabays A, Cao QL, Sugeng L, Magni G, Yao J, Vannan M, Schwartz S, Pandian N. Is there a critical amount of ischemic myocardium required to produce a detectable wall motion abnormality? A quantitative 3-dimensional echocardiographic contrast study. Circulation. 1995;92(suppl I):I-261. Abstract.
Jiang L, Siu SC, Handschumacher MD, Guererro JL, Vazquez de Prada JA, King ME, Picard MH, Weyman AE, Levine RA. Three-dimensional echocardiography: in vivo validation for right ventricular volume and function. Circulation. 1994;89:2342-2350.