Author + information
- Received December 19, 2011
- Revision received April 9, 2012
- Accepted May 3, 2012
- Published online August 1, 2012.
- Victor Mor-Avi, PhD⁎,⁎ (, )
- Chattanong Yodwut, MD⁎,
- Carly Jenkins, PhD†,
- Harald Kühl, MD‡,
- Hans-Joachim Nesser, MD§,
- Thomas H. Marwick, MD†,
- Andreas Franke, MD‡,
- Lynn Weinert, BS⁎,
- Johannes Niel, MD§,
- Regina Steringer-Mascherbauer, MD§,
- Benjamin H. Freed, MD⁎,
- Lissa Sugeng, MD⁎ and
- Roberto M. Lang, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Victor Mor-Avi, University of Chicago, MC5084, 5841 South Maryland Avenue, Chicago, Illinois 60637
Objectives We studied in a multicenter setting the accuracy and reproducibility of 3-dimensional echocardiography (3DE)–derived measurements of left atrial volume (LAV) using new, dedicated volumetric software, side by side with 2-dimensional echocardiography (2DE), using cardiac magnetic resonance (CMR) imaging as a reference.
Background Increased LAV is associated with adverse cardiovascular outcomes. Although LAV measurements are routinely performed using 2DE, this methodology is limited because it is view dependent and relies on geometric assumptions regarding left atrial shape. Real-time 3DE is free of these limitations and accordingly is an attractive alternative for the evaluation of LAV. However, few studies have validated 3DE-derived LAV measurements against an accepted independent reference standard, such as CMR imaging.
Methods We studied 92 patients with a wide range of LAV who underwent CMR (1.5-T) and echocardiographic imaging on the same day. Images were analyzed to obtain maximal and minimal LAV: CMR images using standard commercial tools, 2DE images using a biplane area-length technique, and 3DE images using Tomtec LA Function software. Intertechnique comparisons included linear regression and Bland-Altman analyses. Reproducibility of all 3 techniques was assessed by calculating the percentage of absolute differences in blinded repeated measurements. Kappa statistics were used to compare 2DE and 3DE classification of normal/enlarged against the CMR reference.
Results 3DE-derived LAV values showed higher correlation with CMR than 2DE measurements (r = 0.93 vs. r = 0.74 for maximal LAV; r = 0.88 vs. r = 0.82 for minimal LAV). Although 2DE underestimated maximal LAV by 31 ± 25 ml and minimal LAV by 16 ± 32 ml, 3DE resulted in a minimal bias of −1 ± 14 ml for maximal LAV and 0 ± 21 ml for minimal LAV. Interobserver and intraobserver variability of 2DE and 3DE measurements of maximal LAV were similar (7% to 12%) and approximately 2 times higher than CMR (4% to 5%). 3DE classified enlarged atria more accurately than 2DE (kappa: 0.88 vs. 0.71).
Conclusions Compared with CMR reference, 3DE-derived LAV measurements are more accurate than 2DE-based analysis, resulting in fewer patients with undetected atrial enlargement.
It is well established that increased left atrial volume (LAV) is associated with adverse cardiovascular outcomes and is among the first criteria used to diagnose left ventricular (LV) diastolic dysfunction, according to the current recommendations of the American Society of Echocardiography (1). Today, LAV measurements are routinely performed using 2-dimensional echocardiography (2DE). Most commonly, LAV is estimated using either the single- or bi-plane area-length technique or method of disks (2). The accuracy of these approximations is limited because of their view dependency and their reliance on geometric assumptions regarding left atrial (LA) shape. In addition, LA remodeling as a result of disease processes is frequently asymmetrical, rendering the standard geometric assumptions even more inadequate. Not surprisingly, several studies have shown that 2DE underestimates LAV compared with imaging techniques that are free of these limitations, such as cardiac computed tomography (3,4) and cardiac magnetic resonance (CMR) (5). Importantly, this underestimation may contribute to misclassification of patients with LV diastolic dysfunction (6).
Because real-time 3-dimensional echocardiography (3DE), the natural extension of 2DE, can overcome these limitations by allowing direct detection of LA boundaries in 3-dimensional space, it is attractive as a potentially more accurate and more reproducible alternative for the LAV measurements (7–9). However, few studies have prospectively validated 3DE-derived LAV measurements in large groups of patients against an accepted independent reference standard (10,11) because of the lack of appropriate tools for volumetric analysis of LAV from 3DE datasets, the high cost of CMR or cardiac computed tomography studies, and the radiation concerns associated with the latter. Also, none of the published studies were performed prospectively in a multicenter setting.
Accordingly, this study was designed to: 1) prospectively validate in a multicenter setting new dedicated volumetric software for LAV measurement from 3DE datasets against a CMR reference; 2) compare the accuracy and reproducibility of this analysis with those of 2DE-based LAV measurements; and 3) determine the added value of 3DE over 2DE measurements of LAV for the interpretation of LV diastolic function.
To achieve these goals, we studied patients with a wide range of LAV referred for CMR evaluation at 4 institutions. In each patient, 2DE, 3DE, and CMR imaging were performed on the same day. All images were analyzed to measure LAV at end-ventricular systole (just before mitral valve opening), while excluding the LA appendage. The LAV index (LAVi), defined as LAV normalized by body surface area, was also calculated. The 2DE and 3DE measurements were compared with the CMR reference values. Repeated measurements were performed to assess the reproducibility of LAV measurement by the 3 techniques: 2DE, 3DE, and CMR imaging.
Ninety-two patients (48 ± 18 years of age, 57 men and 35 women, body surface area of 1.72 ± 0.34 ml/m2), referred for clinically indicated CMR for a wide range of suspected cardiovascular conditions, were studied. These 92 patients were selected from 107 consecutive patients, who agreed to participate, on the basis of adequate echocardiographic image quality to allow LAV measurements. CMR-derived maximal LAVs ranged between 40 ml and 206 ml. Exclusion criteria were previous cardiac surgery and known contraindications to CMR, including pacemaker or defibrillator implantation, atrial arrhythmia, claustrophobia, and dyspnea precluding a 10- to 15-s breath-hold. The protocol was approved by the institutional review board of each participating institution. Written informed consent was obtained from each patient.
Magnetic resonance imaging
CMR images were obtained using a 1.5-T scanner with a phased-array cardiac coil. Equipment manufacturers varied among institutions and included Philips (Intera Achievea, Best, the Netherlands), Siemens (MAGNETOM Sonata, Erlangen, Germany), and General Electric (Sigma EXCITE, Milwaukee, Wisconsin). In each patient, retrospective electrocardiogram-gated localizing spin-echo sequences were used to identify the long axis of the heart. Steady-state free-precession dynamic gradient echo (balanced TFE for Philips, TrueFISP for Siemens, and FIESTA for General Electric) cine loops of the left atrium (8-mm thick short-axis slices with 2-mm gaps and 2.0 × 2.0-mm in-plane spatial resolution) were then obtained using retrospective electrocardiographic gating and parallel imaging techniques (SENSE for Philips, mSENSE for Siemens, and ASSET for General Electric) during 10- to 15-s breath-holds with a temporal resolution of 30 frames per cardiac cycle.
Magnetic resonance image analysis
CMR images acquired at each site were analyzed on-site using commercial software supplied by the corresponding manufacturer (Philips ViewForum, Siemens Argus, General Electric MASS Analysis). Analysis included slices from the most proximal slice that showed the LA cavity to the most distal slice that showed the LA cavity with no more than 50% of the circumference surrounded by LV myocardial tissue (12,13). The LA boundary was manually traced in every slice at end-systole (i.e., just before mitral valve opening). All tracings were performed by investigators experienced in CMR analysis who were blinded to echocardiographic measurements. The LAV was calculated using the disk area summation method (modified Simpson rule). These values were used as a reference for comparison with the 2DE and 3DE data.
Transthoracic harmonic imaging was performed using the iE33 system (Philips Medical Systems, Andover, Massachusetts) with the patient in the left lateral decubitus position. Before each acquisition, images were optimized for endocardial visualization by modifying the gain, compress, and time gain compensation controls. 2DE imaging was performed using the S5-1 phased-array transducer in apical 2- and 4-chamber views optimized for maximal LA cavity size, and 3 cine loops were acquired in each view. 3DE imaging was performed using an X3-1 matrix array transducer (Philips Medical Systems) from the apical position. A wide-angled acquisition full-volume mode over 5 consecutive cardiac cycles during a single breath-hold was used. Special care was taken to include the entire LA cavity within the pyramidal 3D volume. Three dynamic pyramidal datasets were acquired.
2DE images collected at all sites were analyzed at the core laboratory at the University of Chicago. The best 2DE cine loops were selected and analyzed at each site using the Xcelera workstation equipped with QLAB software (Philips Medical Systems) by an experienced investigator who was blinded to the results of the CMR and 3DE measurements. In each view (apical 2 and 4 chamber), the LV end-systolic frame corresponding to the largest LA area, just before the mitral valve opening, was selected for analysis. In this frame, LA boundaries were traced manually in both views while excluding the LA appendage and the pulmonic veins. Then, the end-systolic LAV was calculated using the biplane area-length technique (2). In addition, the frame depicting the smallest LA cavity was analyzed in a similar manner to obtain the minimal LAV.
All real-time 3DE images were analyzed at the core laboratory at the University of Chicago. The best real-time 3DE dataset was selected and analyzed offline using prototype software designed for volumetric analysis of the left atrium (Tomtec 3D LA Function, Tomtec Imaging Systems, Unterschleissheim, Germany) by an experienced investigator who was blinded to the results of the CMR and 2DE measurements. First, 2-, 3- and 4-chamber views were automatically selected by the software from the real-time 3DE pyramidal dataset. In each view, the LA boundaries were manually initialized on 2 frames depicting minimal and maximal LAVs (Fig. 1,Online Video 1). During initialization, the blood–tissue interface was traced manually. These initialized LA boundaries were used to reconstruct the LA endocardial surface, the 3D rendering of which was superimposed on 3D gray-scale dataset, allowing visualization from any desired angle (Fig. 2). Reconstruction of the LA endocardial surface was then repeated for each frame throughout the cardiac cycle, resulting in a dynamic cast of LA cavity (Fig. 3, left and center). For each consecutive frame, the voxel count inside the 3-dimensional surface was used to measure the LAV, resulting in a smooth interpolated LAV time curve with effective temporal resolutions of 150 to 200 samples per second (Fig. 3, right) and allowing easy detection of the maximal and minimal LAV.
To determine the reproducibility of the maximal LAV measurements for each imaging modality, CMR, 2DE, and 3DE image analysis was repeated in a randomly selected group of 33 study subjects by an additional investigator as well as by the same primary reader at least 1 week later. In addition, to determine the test-retest reproducibility of 3DE measurements of the maximal LAV in 22 randomly selected patients, analysis was repeated on a different 3DE dataset. During all repeated analyses, the investigators were blinded to the results of all previous measurements.
Impact of imaging modality on the classification of LV diastolic function
The aim of this analysis was to determine in what percentage of patients potentially inaccurate determination of LA size by 2DE and 3DE may contribute to erroneous diagnosis of LV diastolic dysfunction. To achieve this goal, LAVi values obtained by the 3 imaging modalities were used to classify each patient's LA size as normal or abnormal using the abnormality threshold of LAVi >34 ml/m2, as a first step of the American Society of Echocardiography recommendations for grading diastolic dysfunction (1). Then CMR-based determinations of abnormally increased LAVi were used as a reference against which 2DE and 3DE determinations were tested.
2DE- and 3DE-derived values of the LAV were compared with the corresponding CMR reference values using linear regression with Pearson correlation coefficients and Bland-Altman analyses to assess the bias and limits of agreement (defined as 2 SD around the mean) with the CMR reference. To verify the significance of the biases, paired t test versus null values were applied. Values of p < 0.05 were considered significant. Interobserver and intraobserver variability were calculated as an absolute difference in the corresponding pair of repeated measurements in the percentage of their mean in each patient and then averaged over the entire study group. To determine the impact of imaging modality on the classification of LV diastolic function, 2 × 2 contingency tables of normal versus enlarged LAVi were created for 2DE and 3DE data against CMR reference. Kappa statistics were used to assess the accuracy of each technique. The calculated kappa coefficients were judged as follows: 0 to 0.2, low; 0.21 to 0.4, moderate; 0.41 to 0.6, substantial; 0.61 to 0.8, good; and >0.8, excellent.
Comparisons with CMR
Figures 4 and 5⇓ show, side by side, the results of the comparisons between the 2DE and 3DE measurements of the maximal and minimal LAVs, respectively, against the corresponding CMR values. 2DE-derived values of the LAV correlated well with CMR reference values (r = 0.74 and r = 0.82, respectively). However, Bland-Altman analysis revealed negative biases of 31 ml and 16 ml, respectively (both p < 0.001), reflecting a systematic underestimation of the LAV by the 2DE technique. There was a trend toward increased bias in patients with enlarged atria compared with those with normal atrial sizes. The corresponding 3DE measurements resulted in even better correlations with CMR (r = 0.93 and r = 0.88) with only minimal biases of −1 ml (NS) for maximal LAV and 0 ml for minimal LAV. The limits of agreement for the 3DE measurements were considerably tighter than those of the 2DE data (2 SD: 50 ml for 2DE and 28 ml for 3DE for maximal LAV; 2 SD: 64 ml for 2DE and 43 ml for 3DE for minimal LAV).
Table 1 shows the results of the reproducibility analysis of the maximal LAV for CMR, 2DE, and 3DE images. Both interobserver and intraobserver variability were higher for both 2DE- and 3DE-derived volumes than for the CMR measurements. Not surprisingly, for all 3 techniques, the interobserver variability was higher than the intraobserver variability. Importantly, all variability values were <10%, with the exception of the interobserver variability of the 2DE measurements, which was 12%. The test-retest variability was slightly higher than the intermeasurement variability (11 ± 10%).
Impact of imaging modality on the classification of LV diastolic function
Table 2 shows the agreement between 2DE and 3DE classifications of a normal versus enlarged LAV with the CMR reference. 2DE classifications showed agreement with CMR in 65 of 92 patients, corresponding to kappa = 0.71 and reflecting good intertechnique agreement. Of note, the majority of the discordances between 2DE and CMR were false-negative classifications (25 vs. 2 false-positive classifications) (i.e., atria that were enlarged according to CMR were classified as normal when measured by 2DE). 3DE classifications agreed with CMR in 81 of 92 patients, corresponding to kappa = 0.88, indicating excellent agreement. Of note, in contrast to the 2DE classifications, the discordances were more evenly distributed between false-positive and false-negative classifications (4 and 7 patients).
The prognostic value of LAV as a predictor of morbidity and mortality is well established in multiple disease states (14–21) and is of particular importance in the context of the evaluation of LV diastolic function (1,6,13,22). However, the 2DE techniques routinely used in clinical practice to measure LAV are prone to errors because of their 2-dimensional nature. It is true that the rotational symmetry is usually a reasonable assumption for the normal LA shape, as opposed to the frequently asymmetrical left ventricle and that the dependence on the choice of imaging plane is partially alleviated by the use of biplane techniques (23). Nevertheless, abnormally enlarged atria are frequently also asymmetrical, rendering the standard geometric assumptions inadequate. In addition, the accuracy of the biplane measurements is still limited because it is difficult to ascertain that the long axis of the atrium is perfectly aligned with the imaging plane. This is because the LAV is frequently measured in apical views optimized for the left ventricle; in other words, the atrium may be foreshortened because its long axis is not necessarily perfectly aligned with that of the ventricle.
Indeed, several recent studies that have reported that 2DE measurements of LAV are underestimated compared with cardiac computed tomography (3,4) and CMR (5) reference values. There are several publications showing that this underestimation was considerably reduced by the use of 3DE imaging (7,8). However, all previous studies were performed in single-center settings and involved relatively small numbers of patients. Another reason for the relative scarcity of data on 3DE measurement of LAV is the lack of software tools for volumetric analysis applicable to this chamber. As such tools are developed and tested, they need to be validated in a sufficiently large number of patients with a wide range of LAVs. Our study was designed to address this need by testing newly developed software specifically designed for volumetric analysis of the left atrium in a multicenter setting that is better suited to demonstrate general applicability than single-center studies.
The results of our study confirmed that in the majority of patients, biplane measurements of LAV were significantly underestimated compared with CMR values, especially in patients with enlarged atria, despite the fact that in this study we used 2DE apical views optimized for maximal LA size to measure LAV (Fig. 4). As a result, a considerable proportion of these patients were misclassified as having normal size left atria (Table 2). In contrast, volumetric measurements from 3DE datasets resulted in no significant bias (Fig. 4), as one might expect following the above reasoning. Importantly, the 3DE approach showed the same percentage of enlarged atria in our study group as did CMR (86%), and the number of discordant classifications of LAV was reasonably low (14%) and equally distributed between false-positive and false-negative classifications (Table 2).
Although the 3DE approach is an attractive alternative to CMR because of the high speed of acquisition combined with the low cost and widespread use of echocardiographic imaging, our results also showed that this methodology has limitations. Importantly, the intermeasurement variability of 3DE was not better than that of the traditional 2DE measurement, and both were approximately twice as high as that of CMR. Also, despite the fact that 3DE measurement has not shown a systematic error compared with CMR (i.e., significant bias), the limits of agreement were quite wide, indicating that in individual patients the intertechnique differences can be large and that the 2 imaging modalities are certainly not interchangeable.
On the other hand, does this mean that 3DE measurements of LAV are inaccurate? It is important to remember that CMR is not necessarily the perfect gold standard reference technique in this context. As opposed to the excellent in-plane spatial resolution of CMR imaging, this modality is not truly 3-dimensional but rather multiplane, with slice thickness that may be suboptimal for LAV quantification. This is because the number of slices spanning the left atrium can be rather small (as small as 3 or 4 in the case of smaller atria), which is not sufficient to ensure high accuracy when the method of disks is used.
In addition, one might wonder why 3DE-derived LAV measurements are relatively accurate, whereas LV volumes are reportedly underestimated compared with CMR reference. Our previous study (24) demonstrated that the major source of underestimation in 3DE-derived LV volumes is the inability of 3DE to differentiate in every patient between endocardial trabeculae and the myocardium. As a result, the LV endocardial boundary is frequently traced incorrectly, while excluding the trabeculae from the LV cavity. This source of error does not apply to LAV measurements because trabeculae do not exist in the left atrium.
It is important to remember that the American Society of Echocardiography–recommended threshold for LV enlargement (LAVi >34 ml/m2) was derived from 2DE measurements, and thus its use as a reference for what constitutes normal LA volume on CMR, as we did in this study, may be questioned. We used this threshold simply because no other threshold has been established for either CMR or 3DE measurements of LA volume. However, given the systematic differences in values obtained using 2DE compared with 3DE and/or CMR, it is clear that 2DE is likely to misclassify more patients as having normally sized volumes compared with the other 2 imaging modalities. Nevertheless, the important lesson from these comparisons is that technique-specific thresholds for LA enlargement need to be established for 3DE and CMR to optimize their use for the evaluation of LV diastolic function. Another important lesson for laboratories that rely on 2DE measurements of LAV is to interpret their measurements with caution, especially in patients whose measured values are slightly below the abnormality threshold of 34 ml/m2.
One limitation of the 3DE approach used in this study is that it requires manual initialization of the LA boundaries in 6 frames (3 views, 2 frames per view), which takes 3 times longer than the 2DE measurement. Also, the ability to correctly identify LA boundaries depends on the image quality and can be reasonably expected to yield inaccurate measurements in patient with suboptimal acoustic windows. One of the limitations of our study is that we did not compare between imaging modalities the passive and active LA emptying components. This is because in many patients, the prototype analysis software used in this study resulted in incomplete LAV time curves with a portion missing at the end of the cardiac cycle, making it difficult to accurately determine the active versus passive emptying volumes, especially in patients with higher heart rates, in whom the separation between these phases is less obvious.
This is the first prospective multicenter study to validate new dedicated software for volumetric measurement of LAV in a large group of patients with a wide range of LA sizes. Compared with the CMR reference, 3DE-derived LAV measurements were found to be more accurate than 2DE-based analysis with similar reproducibility. Importantly, the improved accuracy resulted in fewer patients with undetected atrial enlargement, which has important implications for echocardiographic evaluation of LV diastolic function.
For a supplementary video and its legend, please see the online version of this article.
Each of the 4 participating sites received a research grant from Philips Healthcare. Dr. Yodwut received a fellowship from the Department of Medicine, Ramathibodi Hospital, Mahidol University, Bangkok, Thailand. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Sherif Nagueh, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- 2-dimensional echocardiography
- 3-dimensional echocardiography
- cardiac magnetic resonance
- left atrial
- left atrial volume
- left atrial volume index
- left ventricular
- Received December 19, 2011.
- Revision received April 9, 2012.
- Accepted May 3, 2012.
- American College of Cardiology Foundation
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Osranek M.,
- Bursi F.,
- Bailey K.R.,
- et al.
- Bangalore S.,
- Yao S.S.,
- Chaudhry F.A.
- Ristow B.,
- Ali S.,
- Whooley M.A.,
- Schiller N.B.
- Lim T.K.,
- Dwivedi G.,
- Hayat S.,
- Majumdar S.,
- Senior R.
- Mor-Avi V.,
- Jenkins C.,
- Kühl H.P.,
- et al.