Background: Real-time 3-dimensional echocardiography (RT3D) is a recent technique based on volumetric scanning, eliminating the need for geometric modeling of the cardiac chambers and minimizing the errors caused by foreshortened views.
Hypothesis: Estimations of left ventricular (LV) end-diastolic (EDV) and end-systolic volume (ESV), and left atrial (LA) size, differ depending on the echocardiographic technique of estimation.
Animals: Fifty-one dogs with acquired heart disease and 34 healthy control dogs.
Methods: Prospective observational study by M-mode (Teichholz method), Simpson's modified 2-dimensional (2D) method, and RT3D methods for estimation of LV volumes. LA size was evaluated by 2D and RT3D methods.
Results: RT3D showed good agreement with 2D for EDV and ESV, whereas Teichholz method overestimated LV volumes in comparison with the other 2 methods by approximately a factor 2. There were no statistically significant differences among the 3 methods in estimating ejection fraction. Comparison between RT3D assessment of LA end-systolic volume per kilogram (LAs/kg) and LA to aortic ratio (LA/Ao) measured by 2D relative to each other showed that the RT3D method underestimated LAs/kg at lower values, and overestimated it at higher values. The difference between methods increased with increasing LA size.
Conclusions and Clinical Importance: There was good agreement between RT3D and 2D methods of estimating EDV and ESV, whereas the Teichholz method overestimated LV volumes by approximately a factor 2. In comparison with RT3D, LA/Ao underestimated LA size, especially when LA was enlarged.
Teichholz method for estimation of volumes from M-mode
Quantification of cardiac chambers, including left ventricular (LV) volumes and left atrial (LA) size, and assessment of LV systolic function are essential parts of the echocardiographic examination. A quantitative assessment of global systolic function usually is based on changes in ventricular size and volume. Previously, only 1-dimensional (M-mode) and 2-dimensional (2D) echocardiography were available for these purposes. More recently, 3-dimensional (3D) techniques have been developed and evidence is accumulating that 3D imaging might be superior to M-mode and 2D techniques in quantification of LV volume and ejection fraction (EF).1 Teichholz and colleagues developed an equation using an ellipsoid model of the LV for estimation of end-systolic (ESV) and end-diastolic (EDV) volumes from M-mode images. Correlations with angiographically derived LV volumes were considered accurate, although caution was recommended for patients with coronary artery disease and LV dyssynchrony.2 Conventional 2D echocardiography provides cross-sectional views of the heart, and different geometric models have been used to estimate ventricular volumes.3 The most widely used method is the modified Simpson's biplane method or disc summation method, where the volume of the LV is calculated from the dimension and area obtained from 2 orthogonal left parasternal apical views, 4-chamber (4Ch) and 2-chamber (2Ch) views. By this method, it is essential to have nonforeshortened apical views with similar long-axis dimensions in both views. Reliable visualization and tracing of endocardial blood-tissue borders are pivotal for accurate LV volume measurements.
3D echocardiography was first described by Dekker et al.4 Real-time 3-dimensional (RT3D) echocardiography was reported in 1990, using a sparse array matrix transducer consisting of 256 nonsimultaneously firing elements.5,6 Images obtained by this system were of relatively poor quality, frame rates were low, the pyramidal volumes were of a relatively narrow sector, and images were not volume-rendered on-line. Presently, technology is available using a matrix transducer with 3000 individual elements generating ultrasonic beams in a phased array manner, which are automatically aimed in multiple directions, allowing simultaneous visualization of the beating heart. 3D data sets are obtained from several cardiac cycles. During each cardiac cycle, a wedge-shaped subvolume is acquired, and the combined subvolumes from 4 cardiac cycles provide a full-volume dataset. With the use of cropping controls, 3D images of the heart can be dissected from a full-volume pyramid, and 3D images can be rotated in any direction for viewing.7 The relative inaccuracy of M-mode and 2D images may be attributed to the need for geometric modeling of chambers and to errors caused by foreshortened views. Hence, the most important advantage of 3D echocardiography is independence of geometric modeling and image plane positioning.1
Assessment of LA size is crucial in the evaluation of heart disease as it allows the examiner to estimate the risk of developing left-sided congestive heart failure (CHF).8 By convention, the size of LA is determined at end-systole or early diastole, either in 2D parasternal short-axis or long-axis views. Different methods of measuring LA diameter, circumference, and cross-sectional area have been used, and correlation is commonly made to aortic (Ao) diameter, circumference, and cross-sectional area, respectively.9 LA diameters and area measurements, especially in enlarged LA, were shown to be poor predictors of 3D LA volume in a study of human patients.10
The purpose of this study was to compare estimates of LV volume and LA size by M-mode, 2D, and RT3D echocardiography in dogs with and without acquired heart disease.
Materials and Methods
Animals and Procedures
Fifty-one dogs presented for evaluation of heart disease at Albano Animal Hospital, Stockholm, Sweden and 34 staff- and client-owned healthy control dogs were recruited for the study. Dogs with congenital heart disease were excluded from the study. All 85 dogs were examined with the same equipment and the same protocol. Dogs were considered healthy based on normal findings on physical examination, ECG, echocardiography, and Doppler examinations. All examinations were performed and evaluated by 1 veterinary specialist in cardiology (A.T.).
M-Mode, 2D, and 3D Echocardiography
M-mode, 2D, and RT3D examinations were performed with an ultrasound unita equipped with 3.0–8.5 MHz phased-array transducers (for 2D and M-mode) and 3 or 7 MHz matrix transducers (for RT3D) in all dogs. Dogs were unsedated and gently restrained in left and right lateral recumbency during the examination. Measurements of LV diameters were made by 2D-guided M-mode according to the American Society of Echocardiography.11 End-diastole was defined as the 1st frame after mitral valve closure and end-systole was defined as the frame before mitral valve opening. Estimations of LV volumes and EF (Teich-EF) were derived from Teichholz method2 in diastole (Teich-EDV) and systole (Teich-ESV). Modified Simpson's biplane method was used to calculate LV volumes using 4Ch and 2Ch images obtained from the left parasternal apical view.12,13 RT3D images of the LV and LA were obtained using the X3 or X7 matrix transducer (depending on the size of the dog) to obtain a pyramidal volume in real time. Four smaller real-time volumes, acquired from alternate cardiac cycles, were combined to produce a larger pyramidal volume, providing a full-volume data set. The lowest possible scan line density was used in all dogs to produce the largest possible pyramidal volume of acquisition. Measurements of Ao and LA in early diastole were made on the 2D parasternal short-axis view obtained at the level of the Ao valve at the 1st frame after Ao valve closure.14 Measurements on M-mode images of LV and 2D images of Ao and LA were made directly on the monitor freeze-frame image. Measurements were made on 3 consecutive cardiac cycles.
Off-line analyses of 2D and RT3D volumes were made by a software program.b A template, defined by 6 reference points placed at the mitral annulus hinge points and at the apex in both 4Ch and 2Ch views, was used for the modified Simpson's method to estimate end-diastolic (2D-EDV) and end-systolic (2D-ESV) LV volumes and EF (2D-EF). The template in each view was manually adjusted to optimize LV volume when necessary. Analyses of LV volumes and EF (RT3D-EF) by RT3D involved manual definition of 4 reference points placed at the endocardial border of the mitral annulus hinge points in both 4Ch and 2Ch views and a 5th reference point at the endocardial border of the apex in either view in both end-diastole (RT3D-EDV) and end-systole (RT3D-ESV). The endocardial border then was traced by an automated detection process to create a cast of the LV cavity (Fig 1). When the volume was computed, the endocardial border detection was verified for accuracy and the endocardial borders including papillary muscles were manually edited when necessary. The same method was used for the analysis of LA volumes, where the 4 reference points were placed on the corresponding LA side of the mitral annulus in both 4Ch and 2Ch views, and a 5th reference point placed at the midpoint of the dorsal LA border at end-systole (LAs) and end-diastole (LAd). Three acquisitions were made for 2D and RT3D datasets. Measurements were averaged for each variable, and mean values were used in the statistical analysis. All measurements were performed by 1 veterinary specialist in cardiology (A.T.).
Assessment of Variability
Within-day variability was assessed using 6 dogs, including 3 dogs without heart disease and 3 dogs with myxomatous mitral valve disease (MMVD) (Class BI according to the CHIEF classification of CHF).15–17 Each dog was examined 6 times on a given day and M-mode, 2D, and RT3D derived volumes of LV and 2D diameters and RT3D derived volumes of LA were estimated. Each variable was measured on 3 consecutive acquisitions and the resulting mean values and standard deviations were used to determine the coefficient of variation (CV) (Table 1). All CV values were below 15% except for 2D-ESV (17.8%) and Teich-ESV (16%), and CV values were generally lower for RT3D measurements (< 11%) compared with 2D measurements and M-mode measurements of LV.
Table 1. Within-day variability of M-mode (Teichholz), 2-dimensional (2D), and real-time 3-dimensional (RT3D) estimations of left ventricular (LV), end-diastolic (EDV), and end-systolic (ESV) volumes and ejection fraction (EF) and RT3D estimations of left atrial volumes in end-diastole (LAd) and in end-systole (LAs) and 2D measurement of LA diameter in 6 dogs (3 dogs without heart disease and 3 dogs with MMVD).
CV (%) and Range
Teichholz LV EDV
Teichholz LV ESV
Teichholz LV EF
Simpson's 2D LV EDV
Simpson's 2D LV ESV
Simpson's 2D EF
RT3D LV EDV
RT3D LV ESV
A computer programc was used for all statistical analyses. Values are reported as medians and interquartile ranges. Kruskal-Wallis test was used for testing equality of medians among the 3 methods. For variables in which the medians were significantly different, a pair-wise comparison between the methods also was performed by Mann-Whitney's U-test with Bonferroni adjustment, in which a P-value <.017 was considered significant. The Bland-Altman plot was used to assess agreement between 2 methods at a time by plotting the mean value of the 2 methods by absolute difference.18 Similarly, a Bland-Altman plot was used to compare LAs volume per kilogram (LAs/kg) estimated by RT3D echocardiography, and LA to Ao ratio estimated by 2D echocardiography. The agreement between the 2 compared methods was further evaluated by fitting a linear curve to the observed points. Estimates of slope of the curve and intercept of the y-axis and their P-values were used to evaluate the presence of systematic differences between the 2 methods. Level of significance was set at P < .05.
Eighty-five dogs of 40 different breeds were included in the study: Newfoundland (11), Cavalier King Charles Spaniel (8), Dachshunds (7), mixed breed (6), Doberman Pinscher (5), Labrador Retriever (4), Australian Kelpie (3), Border Collie (3), American Staffordshire Bullterrier (2), Border Terrier (2), Chinese Crested Powder Puff (2), Golden Retriever (2), Jack Russell Terrier (2), Rottweiler (2), Shetland Sheepdog (2), and 1 each of 24 other breeds. Thirty-nine dogs were diagnosed with MMVD, 12 dogs were diagnosed with dilated cardiomyopathy (DCM), and 34 dogs were healthy controls. Pulmonary hypertension was diagnosed in 4 of the dogs with MMVD and in 1 of the dogs with DCM. According to the CHIEF classification,15–17 34 dogs were classified without CHF (30 in class BI and 4 in BII) and 17 dogs were classified with CHF (15 in class CII and 2 in CIII). Forty-four dogs (52%) were males and 41 dogs (48%) were females. Weight ranged from 4.5 to 60 kg (median, 14.2 kg). Age ranged from 2 months to 15 years (median, 7.3 years). Mean heart rate ranged from 61 to 198 beats/min (median, 118 beats/min). Atrial fibrillation was present in 4 of the dogs with DCM and in 1 dog with MMVD.
M-Mode, 2D-, and RT3D Echocardiography
The RT3D method of estimating EDV showed good agreement with the 2D method with nonsignificant estimates of slope and intercept of the fitted curves in dogs with and without heart disease (Fig 2). The Teichholz method overestimated EDV compared with both RT3D and 2D methods (Table 2), and the difference between methods increased with increasing EDV (Figs 3 and 4). A linear fit between the RT3D and the Teichholz methods yields the following equation:
Table 2. Median values and interquartile ranges of left ventricular end-diastolic (EDV) and end-systolic (ESV) volumes and ejection fraction (EF) estimated with M-mode (Teichholz), 2D (Modified Simpson's method of discs), and real-time 3-dimensional (RT3D) in 85 dogs (39 dogs with MMVD, 12 dogs with DCM, and 34 dogs without heart disease).
Values with different superscript letters indicate statistically significant differences between methods.
Comparing RT3D with 2D estimates of ESV, the RT3D method rendered larger estimates with increasing ESV, as indicated by a significant P-value of the slope of the fitted curve (Tables 3 and 4). The Teichholz method overestimated ESV compared with both RT3D and 2D methods (Table 2), with increasing difference with increasing ESV (data not shown). There was no significant difference between RT3D, 2D, and Teichholz methods of estimating EF in this study (Table 2).
Table 3. Results from linear curve fitting in the Bland-Altman plots in 34 dogs without heart disease comparing 3 different methods, ie, the Teichholz method (Teich), Simpson's 2-dimensional method (2D) and real-time 3-dimensional method (RT3D) of assessing left ventricular end-diastolic volume (EDV), end-systolic volume (ESV), and ejection fraction (EF).
The RT3D method of assessing left atrial volume at end-systole (RT3DLAs) per kilogram (kg) bodyweight was compared with the LA to aortic (Ao) ratio.
Table 4. Results from linear curve fitting in the Bland-Altman plots in 51 dogs with heart disease (39 dogs with MMVD, 12 dogs with DCM) comparing 3 different methods, ie, the Teichholz method (Teich), Simpson's 2-dimensional method (2D), and real-time 3-dimensional method (RT3D) of assessing left ventricular end-diastolic volume (EDV), end-systolic volume (ESV), and ejection fraction (EF).
The RT3D method of assessing left atrial volume at end-systole (RT3DLAs) per kilogram (kg) bodyweight was compared with the LA to aortic (Ao) ratio.
The Bland-Altman plot comparing LAs/kg estimated by RT3D echocardiography and LA to Ao ratio (LA/Ao) estimated by 2D echocardiography showed that the difference between methods increased with increasing LA size. The RT3D method underestimated LAs/kg in comparison with the LA/Ao ratio at lower values, whereas volumes were overestimated at higher values (Fig 6). LA/Ao ranged from 0.7 to 3.82 (median, 1.1) and LAs/kg ranged from 0.3 to 5.8 (median, 0.8). Estimates of LV and LA volumes did not differ significantly between dogs with and without heart disease (Figs 2–5, Tables 3 and 4).
Accurate and reliable estimations of LV volumes form the basis of assessment of LV function and constitute an important part of the echocardiographic examination. RT3D and 2D estimations of LV volumes showed good agreement between methods in this study of dogs with and without acquired heart disease, although both EDV and ESV estimated by RT3D were greater than those estimated by 2D. The Bland-Altman analysis, where the differences between values assessed by 2 methods are plotted against the average of these 2 values, is considered the statistical method best suited for assessing agreement between 2 methods.18 Most studies in humans comparing RT3D and 2D methods show that 2D methods, but not RT3D, consistently underestimate volumes compared with a gold standard technique such as magnetic resonance imaging (MRI).19–21 The Teichholz method overestimated LV volumes compared with both RT3D and 2D techniques in this study. This finding is in agreement with a previous study comparing the Teichholz method and two 2D methods.22 This finding is important because the Teichholz method previously used for calculation of ESV index as an assessment of myocardial function may not be accurate.23 However, from a practical point of view, the Teichholz method is easy to perform with any echocardiographic system, whereas both the RT3D and 2D methods require specific software and commonly off-line analysis. In a clinical situation, the approximate RT3D-EDV can be estimated by dividing the EDV given by the Teichholz method by 2 (Fig 5). It may be argued that the Teichholz method is expected to be more accurate in estimating LV volumes in heart disease, as increasing chamber dilatation and sphericity may render a more cylindrical LV shape, compared with healthy dogs. However, no significant differences were found between dogs with and without heart disease (Figs 2–4, Tables 3 and 4). Estimations of EF by M-mode, 2D, and 3D methods did not significantly differ among methods. Either method thus may be used interchangeably.
Estimation of LA size is pivotal when assessing the risk of developing CHF in an individual animal. Previously, M-mode and 2D methods have been used for the estimation of LA size,9 and increased LA/Ao ratio is considered to be the most important prognostic indicator in dogs with MMVD.24,25 Estimation of LA volume by RT3D was reported to be a major predictor of clinical outcome in human patients with severe LV dysfunction, and its clinical value was superior to that of 2D techniques.26 LA volume also was reported to correlate with the progression of LV diastolic dysfunction.27 Comparison between RT3D measurement of LA volumes and 2D measurement of the LA to Ao ratio in this study of dogs with and without acquired heart disease showed that the difference between methods increased with increasing LA size. The RT3D method underestimated LAs/kg in comparison with the LA/Ao ratio at lower values, whereas volumes were overestimated at higher values. Because there are inherent limitations in using volume estimations on a single-axis dimension, the RT3D method may be assumed to more accurately estimate LA size. However, as the use of the RT3D method requires time-consuming off-line analysis, rapid on-line M-mode or 2D methods still are required in clinical practice. Additional studies are needed to evaluate which M-mode or 2D method of assessing LA size shows best agreement with the RT3D method.
Repeatability was found to be acceptable for most measurements in this study. The RT3D method of estimations of LV variables had the lowest within-day variability of the 3 methods, although all 3 methods showed acceptable repeatability (Table 1). Both RT3D and 2D methods of assessing LA size showed good repeatability (ie, <10%).
In conclusion, this study showed good agreement between RT3D and 2D estimations of LV volumes, whereas the Teichholz method overestimated LV volumes in comparison with the other 2 techniques by approximately a factor 2. LA to Ao ratio underestimated LA size, especially when LA was enlarged. As RT3D estimations of volumes were not evaluated against a gold standard technique such as MRI, this new technique could not be evaluated for accuracy.
aiE33, Philips Ultrasound, Bothell, WA
bQLAB advanced quantification, version 5.0, Philips Ultrasound