Two-dimensional (2D) speckle tracking–derived left atrial (LA) strain (ϵ) facilitates comprehensive evaluation of LA contractile, reservoir, and conduit function; however, its dependence on the individual software used for assessment has not been evaluated. The aim of this study was to compare LA ϵ derived from two different speckle-tracking software technologies, Velocity Vector Imaging (VVI) and 2D speckle-tracking echocardiography (STE).
VVI-derived and 2D STE–derived global longitudinal LA ϵ and ϵ rate (SR) were directly compared in 127 patients (mean age, 62 ± 10 years) with atrial fibrillation. Peak negative, peak positive, and total ϵ (corresponding to LA contractile, conduit, and reservoir function) were measured during sinus rhythm. Late negative (LA contraction), peak positive (left ventricular systole), and early negative (left ventricular early diastole) SR were also measured.
The measurement of LA ϵ and SR by both software was feasible in high proportions of patients (93% with VVI and 93% with 2D STE). The average analysis of ϵ negative was −7.24 ± 3.87% by VVI and −7.30 ± 3.37% by 2D STE ( P = .84). The average analysis of ϵ positive was 14.52 ± 5.82% by VVI and 10.74 ± 4.51% by 2D STE ( P < .01). The average analysis of ϵ total was 21.76 ± 7.39% by VVI and 18.04 ± 5.98% by 2D STE ( P < .01). VVI-derived and 2D STE–derived ϵ positive , ϵ negative , and ϵ total had good correlations with one another ( R = 0.79, R = 0.75, and R = 0.80), with low mean differences. Late negative, peak positive, and early negative SR were correlated less well ( R = 0.78, R = 0.71, and R = 0.67).
LA ϵ measurement using both VVI and 2D STE is feasible in a large proportion of patients in clinical practice. VVI and 2D STE provide comparable LA ϵ and SR measurements for LA contractile function.
A resurgence of interest in atrial function has enhanced our understanding of the atrial contribution to cardiovascular performance in health and disease. Assessment of left atrial (LA) function has evolved over the years. In 1995, Grimm et al. proposed qualitatively assessed increased LA appendage stunning after cardioversion as a possible mechanism of embolization after cardioversion. Ten years later, with the help of technological advances, Di Salvo et al. were able to show that patients with quantitative high LA strain (ϵ) are more likely to maintain sinus rhythm after cardioversion. The quantitative assessment of LA function can be particularly important, because the LA contribution to left ventricular (LV) filling is dynamic and evolves during the course of heart failure. Previous work has shown a restrictive filling pattern to be an independent predictor of adverse events. Evaluation of LA function may emerge as an important component in assessing the hemodynamic effects of a number of diseases, such as atrial arrhythmias, heart failure, and mitral valve disease. The number of drugs, devices, ablative procedures, and surgeries available for the treatment of atrial fibrillation is increasing. Atrial distensibility and contraction both before and after ablation have a clinical impact in patients with atrial fibrillation to assess the thromboembolic risk, procedural success of ablation, structural remodeling after ablation, and atrial fibrosis.
The components of LA function (reservoir, conduit, and contractile function) are traditionally estimated by Doppler analysis of transmitral and pulmonary vein flows. However, the evaluation of LA function can be challenging and warrants the use of sophisticated tools such as two-dimensional (2D) speckle tracking. Recently, we reported that it was possible to evaluate the three components of LA function using 2D speckle-tracking echocardiography (2D STE; GE Medical Systems, Milwaukee, WI), and the values provided were correlated with traditional echocardiographic indexes used to evaluate LA reservoir, conduit, and contractile function. This analysis may allow a more direct assessment of LA contractility, and passive deformation has been shown to change with treatment, such as cardiac resynchronization therapy. This direct assessment of LA function using 2D speckle tracking is likely to overcome the limitations of indirect conventional methods.
Various speckle-tracking software packages are available that differ primarily in their tracking algorithms. To track speckles and compute ϵ and ϵ rate (SR) values, filtering algorithms are used. The effect of this filtering on the results represents a “black box” and may vary from vendor to vendor. It is thus unclear how values from different scanners and software versions compare. Velocity Vector Imaging (VVI; Siemens Medical Solutions USA, Inc., Mountain View, CA) and 2D STE are two of the most commonly used techniques. These techniques quantify deformation by measuring lengthening and shortening relative to the baseline (Lagrangian ϵ). However, the availability of different software technologies to measure this ϵ from different vendors has created the need for a comparative study to assess the agreement between these techniques. In the present study, we sought to determine not only the feasibility of the currently available speckle-tracking echocardiographic ϵ techniques (2D STE and VVI) but also the agreement between the two techniques.
We studied 127 patients in normal sinus rhythm who underwent echocardiography using a GE machine within a time period of no more than 6 months before or after pulmonary vein isolation for atrial fibrillation from April 1 to November 30, 2010, at Cleveland Clinic Heart and Vascular Institute. The major reasons for excluding patients from participation were the absence of an echocardiographic exam using a GE machine within the 6 months from pulmonary vein isolation ( n = 252), atrial fibrillation at the time of examination ( n = 125), and severe valvular heart disease ( n = 12). The study was approved by the Institutional Review Board at Cleveland Clinic.
In all patients, conventional transthoracic M-mode, 2D, and Doppler echocardiography was performed in the left lateral decubitus position. An ultrasound imaging system (Vivid-E9; GE Medical Systems) was used with an M3S phased-array transducer. Cardiac dimensions were measured in accordance with the recommendations of the American Society of Echocardiography. Echocardiograms were stored digitally and reviewed offline with software (syngo Dynamics 9.0; Siemens Medical Solutions USA, Inc.). The values for 2D echocardiographic parameters were obtained after averaging three consecutive cycles. Standard M-mode measurements were taken. The LV ejection fraction was measured using the biplane modified Simpson’s method (using apical two-chamber and four-chamber views). LA volume was determined using the modified Simpson’s method from apical two-chamber and four-chamber views at end–ventricular systole.
Pulsed Doppler echocardiography of transmitral flow was performed. The sample volume was positioned at the level of the mitral tips (transmitral flow) in the apical four-chamber view. Images were recorded at a speed of 50 to 100 mm/sec. From transmitral recordings, the peak early (E) and late (A) diastolic filling velocities, E/A ratio, and E-wave deceleration time were obtained.
An apical four-chamber view was also used to obtain Doppler tissue imaging of the mitral annulus. Two-dimensional pulsed tissue Doppler recordings with second-harmonic imaging were collected with a mean frame rate of 142 frames/sec during brief apnea after expiration. The pulse repetition frequency was adjusted to avoid aliasing. Sample volumes were placed on the septal and lateral sides of the mitral annulus. Values shown for systolic (s′), peak early (e′), and late (a′) diastolic annular velocities are averages of the values obtained at septal and lateral positions.
Only clips with good-quality images from apical views, enough depth to include the whole left atrium, and acquired at high frame rates were used for analysis. The average frame rate of the clips for LA ϵ analysis was 66.3 ± 5.9 frames/sec. We used the onset of the P wave as the reference point for the calculation of LA ϵ and SR, as previously proposed. This point was used because it most relevantly represents the LA cavity just before its contraction. The use of the P wave as the reference point enabled the recognition of peak positive global LA ϵ (ϵ positive ), which corresponded to LA conduit function; peak negative global LA ϵ (ϵ negative ), which corresponded to LA contractile function; and the sum of these values, total global LA ϵ (ϵ total ), which corresponded to LA reservoir function. Similarly, we identified peak negative global LA SR (SR late negative ) during LA contraction, peak positive global LA SR (SR positive ) at the beginning of LV systole, and peak negative global LA SR (SR early negative ) at the beginning of LA diastole. LA ϵ and SR 5,11 and they were measured using both techniques.
Measurement of Global Longitudinal LA ϵ with 2D STE
Measurements were performed offline using dedicated software (EchoPAC-PC version 6.0; GE Medical Systems). One cardiac cycle was selected for each view, and the endocardial border was traced manually. A region of interest (ROI) was manually adjusted to include the entire LA wall thickness. Care was taken to avoid including the pericardium in the ROI. The software then selected stable speckles within the LA wall and tracked these speckles frame by frame throughout the entire cardiac cycle. The software then divided the entire LA circumference in up to six conventional segments and provided a tracking quality (green = good, red = bad) for each segment. The adequacy of tracking was verified manually, and the ROI was readjusted to achieve optimal tracking. Segments having bad tracking were excluded from subsequent analysis. The automated software then generated traces depicting regional longitudinal ϵ for each segment, and ϵ values were measured from these traces. The software calculated average ϵ and SR values for six LA segments for each apical view ( Figure 1 ), and the LA ϵ and SR values for each view were the averages of the values obtained for the LA segments at each view, excluding the three LA segments of the anteroseptal wall of the three-chamber view. The final global longitudinal LA ϵ and SR values were the averages of the values of all apical views.
Measurement of Global Longitudinal LA ϵ with VVI
VVI ϵ measurements were performed using customized software (Siemens Medical Solutions USA, Inc.). For each view, endocardial borders were manually traced in the end-systolic frame, and the software subsequently traced the borders in the other frames automatically. The vectors of the velocities of the endocardial and epicardial points were then displayed and then overlaid onto the B-mode images. In segments with poor tracking (assessed subjectively), endocardial borders were readjusted until better tracking was achieved. If this was unattainable, that segment was excluded. Graphical displays of deformation parameters for each segment were then generated automatically and were used for measurement of LA ϵ values. Average ϵ and SR values were calculated as described above.
Interobserver and intraobserver variability for global LA ϵ (negative, positive, and total) were examined. Measurements were performed in a group of 15 randomly selected subjects by one observer then repeated two separate times by two investigators who were unaware of the other’s measurements and of the study time point. Linear regression analysis with Pearson’s correlation coefficient was performed to measure the strength of the relation of LA ϵ between two measurements. The bias (mean difference) and limits of agreement (1.96 standard deviations of difference) between the first and second measurements were determined. Intraclass correlation coefficients were also assessed to evaluate the reproducibility.
Continuous variables are expressed as mean ± SD. Student’s t test was used to compare the LA ϵ and SR variables between VVI and 2D STE. A box plot was used to identify outliers of LA ϵ and SR measurements. Points that were beyond the quartiles by 1.5 interquartile range were deemed to be outliers. Linear regression analysis with Pearson’s correlation coefficient was performed to measure the strength of the relation of LA ϵ and SR between any two methods. Linear regression models were used to calculate the predicted LA ϵ and SR values on 2D STE from VVI LA ϵ measurements. The Bland-Altman method was used to evaluate the bias and limits of agreement between any two methods. P values < .01 were considered statistically significant; Bonferroni’s correction was used for multiple comparisons.
We evaluated 127 patients, whose characteristics are summarized in Table 1 . The mean age of the study patients was 62 ± 10 years, and 69% were men. Subjects presented with normal blood pressures and heart rates. Of this population, 60% had paroxysmal atrial fibrillation and 65% had undergone pulmonary vein isolation.
|Age (y)||62 ± 10|
|Body mass index (kg/m 2 )||29.3 ± 5.9|
|Heart rate (beats/min)||68 ± 14|
|Systolic blood pressure (mm Hg)||123 ± 17|
|Diastolic blood pressure (mm Hg)||74 ± 10|
|Paroxysmal/persistent AF||76 (60%)/51 (40%)|
|Post PVI||82 (65%)|
|AF duration (mo)||65 ± 70|
|Diabetes mellitus||18 (14%)|
|CHADS 2 score||1.07 ± 1.04|
Standard and Doppler Echocardiographic Parameters
Two-dimensional and Doppler echocardiographic features are described in Table 2 . Although overall LV function was preserved, the left atrium was generally dilated, and LV wall thickness was increased (septal wall thickness, 1.2 ± 0.2 cm; posterior wall thickness, 1.1 ± 0.2 cm).
|LA diameter (cm)||4.5 ± 3.4|
|LV internal dimension in diastole (cm)||4.8 ± 0.7|
|LV internal dimension in systole (cm)||3.3 ± 0.7|
|LV ejection fraction (%)||57 ± 10|
|LA area (cm 2 )||22 ± 6|
|Maximum LA volume index (mL/m 2 )||34 ± 14|
|E (m/sec)||0.89 ± 0.32|
|A (m/sec)||0.66 ± 0.92|
|E/A ratio||1.8 ± 0.9|
|E-wave deceleration time (msec)||216 ± 68|
|s′ average (cm/sec)||7.0 ± 1.6|
|e′ average (cm/sec)||8.3 ± 2.1|
|a′ average (cm/sec)||6.7 ± 2.7|
The measurement of LA ϵ and SR by both software packages was feasible in high proportions of patients (93% with VVI and 93% with 2D STE). The average analyses of LA ϵ and SR variables are summarized in Table 3 ; none of the variables were significantly different after adjusting the α value for multiple comparisons. LA ϵ negative and SR late negative showed quite similar results between VVI and 2D STE. Box plots of these variables are shown in Figure 2 . Nine patients had outliers in LA ϵ and SR variables. VVI-derived and 2D STE–derived ϵ positive , ϵ negative , and ϵ total had good correlations with one another ( Figure 3 ). Still, good correlations were obtained after the exclusion of these outliers ( R = 0.80, R = 0.66, and R = 0.71). Bland-Altman analysis revealed good agreement between VVI-derived and 2D STE–derived ϵ negative , with a mean difference of 0.045%, although VVI methods showed slightly higher values than 2D STE–derived ϵ positive and ϵ total , with mean differences of 3.83% and 3.79% and limits of agreement (±1.96 SDs) of ±7.61% and ±8.75% ( Figure 3 ). LA ϵ negative by VVI showed a marginal directional measurement bias, with a linear regression coefficient of R = 0.23 ( P < .05); however, VVI-derived LA ϵ positive and ϵ total showed a significant directional measurement bias, with linear regression coefficients of R = 0.36 ( P < .0001) and R = 0.33 ( P < .001), respectively, indicating higher values at larger LA ϵ ( Figure 3 ).
|LA ϵ negative||−7.24 ± 3.87||−7.30 ± 3.37||.8396|
|LA ϵ positive||14.52 ± 5.82||10.74 ± 4.51||.0037|
|LA ϵ total||21.76 ± 7.39||18.04 ± 5.98||.0043|
|LA SR late negative||−0.84 ± 0.48||−0.79 ± 0.38||.5567|
|LA SR positive||1.02 ± 0.36||0.88 ± 0.30||.0062|
|LA SR early negative||−0.96 ± 0.43||−0.83 ± 0.38||.0093|