Assessment of Left Ventricular Dyssynchrony with Real-time 3-Dimensional Echocardiography: Comparison with Doppler Tissue Imaging

ORIGINAL ARTICLES

Assessment of Left Ventricular Dyssynchrony with Real-time 3-Dimensional Echocardiography: Comparison with Doppler Tissue Imaging Masaaki Takeuchi, MD, Avrum Jacobs, MD, Lissa Sugeng, MD, Tomoko Nishikage, BS, Hiromi Nakai, BS, Lynn Weinert, BS, Ivan S. Salgo, MD, and Roberto M. Lang, MD, Osaka, Japan; Chicago, Illinois; and Andover, Massachusetts

We studied the usefulness and reproducibility of realtime 3-dimensional (3D) echocardiography (RT3DE) for evaluating left ventricular (LV) dyssynchrony, and compared its results with Doppler tissue image (DTI) indices. Full-volume RT3DE data sets and 2-dimensional DTI from apical window were obtained in 122 participants. Using fast 3D border detection software, time to minimum systolic volume (Tmsv) was semiautomatically calculated in each region from a 17-segment model. Several dyssynchrony indices were then calculated: Tmsv-16SD, the SD of Tmsv in 16 of 17 segments, excluding the apical cap; Tmsv-12SD, the SD of Tmsv of 6 basal and 6 middle segments; and Tmsv6SD, the SD of Tmsv of 6 basal segments. These dyssynchrony indices of RT3DE were then compared with two dyssynchrony indices measured by DTI: time to peak systolic velocity (TTPV)-12SD, the SD of time to peak systolic velocity of 12 LV segments; and time to cross over point of temporal axis (TTCO)-12SD, the SD

Cardiac resynchronization therapy (CRT) is a novel

and promising nonpharmacologic therapy for patients with advanced heart failure and wide QRS morphology.1-5 According to the current functional and electrocardiographic (ECG) eligibility criteria, however, one third of the patients receiving CRT have shown no symptomatic improvement and may actually develop adverse left ventricular (LV) remodeling, resulting in further severe symptomatic impairment.1,6,7 In addition, LV mechanical dyssynchrony has been proven to occur even in patients From the Department of Cardiology and Internal Medicine, Tane General Hospital, Osaka, Japan (M.T., T.N., H.N.); Noninvasive Cardiac Imaging Laboratory, Section of Cardiology, Department of Medicine, University of Chicago Medical Center, Chicago, Illinois (A.V., L.S., L.V., R.M.L.) ; and Philips Medical Systems, Andover, Massachusetts (I.S.S.). Reprint requests: Masaaki Takeuchi, MD, Department of Cardiology and Internal Medicine, Tane General Hospital, 1-2-31 Sakaigawa, Nishi-ku, Osaka, 550-0024, Japan (E-mail: [email protected]). 0894-7317/$32.00 Copyright 2007 by the American Society of Echocardiography. doi:10.1016/j.echo.2007.05.001

of time to crossover point of temporal axis. RT3DE data was quantitatively analyzed in 117 of 122 patients. Tmsv-16SD (35 ⴞ 34 milliseconds) was significantly longer compared with Tmsv-12SD (27 ⴞ 30 milliseconds, P < .001) or Tmsv-6SD (23 ⴞ 28 milliseconds, P < .001). Tmsv-16SD increased significantly with the severity of LV systolic dysfunction. Fair correlation was noted among TTPV-12SD, TTCO-12SD, and Tmsv16SD (r ⴝ 0.71, r ⴝ 0.73) and between Tmsv-16SD and LV ejection fraction (r ⴝ 0.80). Concordance rate between TTPV-12SD and Tmsv-16SD for detecting LV dyssynchrony was 79%. The corresponding value between TTCO-12SD and Tmsv-16SD was 80%. In conclusion, Tmsv-16SD correlated well with DTI-derived LV dyssynchrony indices. In addition to LV remodeling, fast border detection RT3DE provides useful parameters for evaluating LV dyssynchrony. (J Am Soc Echocardiogr 2007;20:1321-1329.)

with preserved QRS duration.8,9 Echocardiographic assessment of LV dyssynchrony has been recently shown to be an excellent noninvasive tool for selecting patients with drug-refractory heart failure who would benefit from CRT.5,6,10-15 Several parameters derived from Doppler tissue imaging (DTI) have been reported to correlate with favorable outcomes after CRT. Among them, the SD of the time to peak systolic velocity (TTPV) assessed with DTI in 12 LV segments has been shown to be a promising predictor of LV reverse remodeling.5,15 However, this index is calculated from the basal and mid segments of the 3 standard 2-dimensional apical views, and consequently does not reflect the motion pattern of all LV segments in 3-dimensional (3D) space. In addition, because of the limitation of angle dependency of DTI, this index does not contain information regarding the apical segments. Furthermore, each segment must be evaluated sequentially and, thus, is subject to beat-to-beat variability. In contrast, real-time 3D echocardiography (RT3DE) allows simultaneous assessment of the entire LV from a single full-volume data set acquired from the apex. In previous studies, RT3DE imaging was lim-

1321

Journal of the American Society of Echocardiography December 2007

1322 Takeuchi et al

ited by cumbersome and time-consuming quantification of LV remodeling by measuring volumes and mass, which relies on manual tracing of endocardial borders or semiautomated border detection of multiple 2-dimensional long-axis planes extracted from the volumetric data. This offline analysis method also required geometric interpolation to translate the multiplanar measurements into volumes.16,17 An improvement to this approach–rapid detection of the entire 3D endocardial surface– has been recently developed, allowing direct quantification of regional LV volumes without the need for multiplane tracing.18-20 This semiautomated software uses a 17segment model of the heart to provide regional volume versus time curves, which can be used to assess LV dyssynchrony. The aim of this study was to: (1) determine feasibility of LV dyssynchrony indices acquired with RT3DE; and (2) compare values derived from RT3DE with different DTI indices when acquired in the same patients.

METHODS Study Participants A total of 122 participants, including 21 control subjects and 101 patients (mean age: 61 ⫾ 7 years; 83 men) with a wide range of LV ejection fraction (EF) (13%-81%) were studied. Patients were selected based on image adequacy, ie, patients were excluded if the endocardium was not seen in two contiguous segments. Informed consent was obtained in all patients. Patients were excluded if they had atrial fibrillation or other rhythm disturbances that precluded acquisition of adequate 3D data sets or multiplane tissue Doppler data sets. RT3DE Harmonic real-time 3D imaging was performed using a matrix-array transducer (X3-1, 1.9/3.8 MHz) connected to a commercially available ultrasound machine (iE33, Philips, Andover, Mass) to obtain a pyramidal volume data set from the apical transducer position. Gain and compression controls, and time gain compensation settings, were optimized to enhance image quality. Care was taken to include the entire LV cavity within the pyramidal volume scan. Real-time 3D data sets were acquired using a wide-angle acquisition (93 ⫻ 80 degrees) mode in which 4 wedge-shaped subvolumes (93 ⫻ 20 degrees each) were obtained from 4 consecutive cardiac cycles. Data were acquired from the apical 4-chamber position during held end expiration. Acquisition was triggered to the R wave on the ECG of every cardiac cycle, resulting in a total acquisition time of 4 heartbeats. Data were subsequently transferred to an offline computer for analysis using

commercially available software (3DQ ADV, QLAB, Version 4.2, Philips). First, nonforeshortened, enddiastolic apical 2- and 4-chamber view cut planes semiautomatically were obtained from the pyramidal data set as described previously.21 Then, 5 anatomic landmarks were manually initialized, including two points to identify the mitral valve annulus in each of the two apical views (apical 2and 4-chamber views) and one point to identify the apex in either view. After manual identification of these points, the program automatically identified the 3D endocardial surface using a deformable shell model (Figure 1).22 Adjustments to the automatic surface detection were also performed, if required. Thereafter, the end-systolic frame was selected by identifying the frame with the smallest LV cavity cross-sectional area in both apical views. After initialization, surface detection was then repeated on this frame to obtain end-systolic volumes. Finally, the computer algorithm automatically defined and traced the endocardial border in all frames of the cardiac cycle. The entire process takes approximately 60 seconds and can be done online or offline. “Casts” of the LV endocardium were then automatically obtained from which global LV volumes versus time curves were derived. From these curves, endsystolic volume, end-diastolic volume, and global EF (calculated as: end-diastolic volume ⫺ end-systolic volume/end-diastolic volume ⫻ 100%) were computed. The LV was divided into 17 segments from apex to base according to American Heart Association and American Society of Echocardiography segmentation schema, and curves depicting regional volumes over the cardiac cycle were obtained for each segment. From these regional volume curves (excluding the apical cap, segment 17), the regional ejection time, defined as the time interval between the R wave and minimal end-systolic volume (Tmsv), was automatically calculated. To assess systolic synchrony, the SD of the regional volume time curves was obtained using 16 segments (Tmsv-16SD); 12 segments, 6 basal and 6 middle segments (Tmsv12SD); and 6 basal segments (Tmsv-6SD) in each patient. In addition to the Tmsv indices, the maximal difference of Tmsv was calculated as well, generating the following additional indices of dyssynchrony: the maximal difference or range of Tmsv among 16 segments, among the 6 basal and 6 middle segments; and among the 6 basal segments. Finally, the difference of Tmsv between the basal septal and basal lateral segments and the difference of Tmsv between the basal septal and basal posterior segments were calculated as well.17 DTI DTI was acquired from the standard long-axis LV apical views as previously described.5,7,8 The 2-dimensional echocardiography with DTI color imag-

Journal of the American Society of Echocardiography Volume 20 Number 12

Takeuchi et al 1323

Figure 1 Example of anatomically corrected apical 4-chamber (A), 2-chamber (B), and short-axis (C) views at end-diastolic frame with automatic tracking of endocardial contour (yellow lines). D, Threedimensional reconstruction of left ventricular endocardial cast.

ing was performed using a broadband transducer (S5-1, 1-5 MHz, iE33). Gain settings, filters, and pulse repetition frequency were adjusted to optimize color saturation, and sector size and depth were set to obtain the highest possible frame rate. A minimum of 3 consecutive beats were stored, and images digitized and transferred to an offline computer (SQ, QLAB, Version 4.2, Philips). Myocardial pulsed Doppler velocity profile signals were reconstituted offline from the DTI color images. Regional myocardial velocity curves were obtained from the apical 4-chamber, 2-chamber, and long-axis views. A region of interest (ROI) was placed in the 6 basal and 6 mid segments to obtain tissue Doppler velocity curves, which were averaged from 3 or 4 consecutive beats. From the individual regional velocity curves of the 12 segments, the following time intervals were measured: (1) TTPV; and (2) time to the crossover point in the temporal axis (TTCO). The following indices were then calculated and used to analyze LV dyssynchrony: (1) SD of TTPV of 6 basal and 6 middle segments (TTPV-12SD); (2) SD of TTCO of 6 basal and 6 middle segments (TTCO-12SD); (3) maximal difference of TTPV among 6 basal and 6

middle segments; and (4) maximal difference of TTCO among 6 basal and 6 middle segments. Intraobserver and Interobserver Variability Intraobserver variability was determined by having one observer remeasure these variables in 8 randomly selected participants 1 month apart. Interobserver measurement variability was determined by having a second observer measure LV dyssynchrony indices by RT3DE and DTI data in 8 participants. Intraobserver and interobserver variabilities were calculated as the absolute difference between the corresponding repeated measurements as a percent of their mean. Statistical Analysis Data are expressed as mean ⫾ SD. Continuous data obtained with RT3DE and DTI were compared using the t test or one-way analysis of variance (ANOVA) as appropriate. The ␹2 of Fisher’s exact test analysis was performed for categorical variables. Linear regression analysis was performed to determine the correlation between RT3DE-derived dyssynchrony

Journal of the American Society of Echocardiography December 2007

1324 Takeuchi et al

Table 1 Clinical characteristics, electrocardiographic findings, and global left ventricular functional indices in study participants Age Male Clinical diagnosis DCMP ICM Diabetic CM HCM VHD CAD HHD Others Healthy volunteer HR, beat/min QRS duration, ms LVH by ECG LBBB LVEDV, mL LVESV, mL LVEF, %

61 ⫾ 7 (range: 19-88) 83 (68%) 15 20 9 1 7 19 15 15 21 68 ⫾ 13 (range: 43-115) 108 ⫾ 26 (range: 70-220) 32 (26%) 12 (10%) 106 ⫾ 46 (range: 17-256) 58 ⫾ 44 (range: 7-200) 51 ⫾ 19 (range: 13-81)

CAD, Coronary artery disease; CM, cardiomyopathy; DCMP, idiopathic dilated CM; ECG, electrocardiography; HCM, hypertrophic CM; HHD, hypertensive heart disease; HR, heart rate; ICM, ischemic CM; LBBB, left bundle branch block; LVEDV, left ventricular (LV) end-diastolic volume; LVEF, LV ejection fraction; LVESV, LV end-systolic volume; LVH, left ventricular hypertrophy; VHD, valvular heart disease.

milliseconds; maximal difference or range of Tmsv among 16 segments, 131 ⫾ 133 milliseconds, ANOVA, P ⬍ .001). LV dyssynchrony indices derived from RT3DE demonstrated significant differences between patients with normal LV systolic function and those with LV dysfunction (Figure 3). These differences were more obvious using Tmsv16SD. Specifically, participants with normal LV systolic function (LVEF ⬎ 60%) had uniform LV segmental function (Tmsv-16SD: 15 ⫾ 10 milliseconds). The Tmsv-16SD increased significantly with the severity of LV systolic dysfunction (LVEF 50%-60%, 19 ⫾ 11 milliseconds; LVEF 30%-49%, 37 ⫾ 31 milliseconds; LVEF ⬍ 30%, 82 ⫾ 33 milliseconds, ANOVA, F ⫽ 55.1, P ⬍ .001). When using a cut-off value of Tmsv-16SD greater than 36 milliseconds (⫹2SD of the value obtained in healthy volunteers) to define significant mechanical dyssynchrony, 23 of 25 patients with LVEF less than 30%, 9 of 23 patients with 30% to 50% of LVEF, 2 of 25 patients with 50% to 60% of LVEF, and 2 of 44 patients with greater than 60% of LVEF showed LV dyssynchrony. A good curvilinear correlation was also noted between Tmsv-16SD and LVEF by RT3DE (r ⫽ 0.80, P ⬍ .001). DTI

indices and DTI-derived dyssynchrony indices. A significance level less than .05 was considered statistically significant.

RESULTS Mechanical Dyssynchrony The RT3DE semiautomatic algorithm adequately tracked the endocardial border in 117 of 122 participants. Table 1 shows clinical characteristics, ECG findings, and 3D-derived global LV functional indices in study participants. Figure 2 shows two examples obtained in two patients of RT3DE-derived LV dyssynchrony indices and regional volume curves obtained in 16 segments. In all participants, Tmsv-16SD (35 ⫾ 34 milliseconds) was significantly longer compared with Tmsv-12SD (27 ⫾ 30 milliseconds, P ⬍ .001) and Tmsv-6SD (23 ⫾ 28 milliseconds, P ⬍ .001). The same tendency was also observed (Tmsv6SD, 12 ⫾ 11 milliseconds; Tmsv-12SD, 13 ⫾ 9 milliseconds; Tmsv-16SD, 17 ⫾ 10 milliseconds; Tmsv-6SD vs Tmsv-16SD, P ⬍ .005; Tmsv-12SD vs Tmsv-16SD, P ⬍ .01) when participants were confined to 21 healthy volunteers. Similarly, the maximal difference of Tmsv was lengthened significantly with the increasing number of segments used in the analysis (maximal difference among the 6 basal segments, 57 ⫾ 68 milliseconds; maximal difference among the 6 basal and 6 middle segments, 86 ⫾ 95

With DTI, the mean value of TTPV-12SD (47 ⫾ 34 milliseconds) in all participants was significantly longer than that of TTCO-12SD (30 ⫾ 27 milliseconds, P ⬍ .001). Maximal difference of TTPV among 6 basal and 6 middle segments was also significantly longer compared with the maximal difference of TTCO among 6 basal and 6 middle segments by DTI. The LV dyssynchrony indices derived from DTI also demonstrated significant differences between patients with normal LV systolic function and those with LV dysfunction. Specifically, participants with normal LV systolic function (LVEF ⬎ 60%) had highly uniform LV segmental tissue Doppler performance (TTCO-12SD: 15 ⫾ 8 milliseconds). TTCO12SD increased significantly with the severity of LV systolic dysfunction (LVEF 50%-60%, 22 ⫾ 14 milliseconds; LVEF 30%-49%, 37 ⫾ 29 milliseconds; LVEF ⬍ 30%, 61 ⫾ 32 milliseconds, ANOVA, F ⫽ 25.8, P ⬍ .001) (Figure 4). When using a cut-off value of TTCO-12SD ⫽ 30 milliseconds (⫹2SD of the value obtained in healthy volunteers) to diagnose dyssynchrony, 19 of 24 patients with LVEF less than 30%, 13 of 23 patients with 30% to 50% of LVEF, 6 of 25 patients with 50% to 60% of LVEF, and 2 of 43 patients with greater than 60% of LVEF showed LV dyssynchrony. Although the same tendency was observed when using TTPV-12SD, the differences between groups were less obvious with overlap between groups (participants with LVEF ⬎ 60%, 37 ⫾ 15 milliseconds; LVEF 50%-60%, 28 ⫾ 14 milliseconds; LVEF 30%-49%, 55 ⫾ 35 milliseconds; LVEF ⬍

Journal of the American Society of Echocardiography Volume 20 Number 12

Takeuchi et al 1325

Figure 2 Example of left ventricular (LV) dyssynchrony indices automatically acquired with real-time 3-dimensional echocardiography (RT3DE), LV cast and volumetric regional curve of 16 segments are shown from two patients with dilated cardiomyopathy paced with biventricular pacer. LV ejection fraction by RT3DE is 30% (A) and 22% (B), respectively. Note homogeneous volumetric profile curve and low values of LV dyssynchrony index (A) and heterogeneous volumetric profile and high values of LV dyssynchrony index (B).

30%, 80 ⫾ 47 milliseconds, ANOVA, F ⫽ 16.2, P ⬍ .001). Correlation Between 3D Indices and DTI Indices Correlation between 3D-derived dyssynchrony indices and DTI-derived indices are shown in Table 2. Correlation was improved with the increasing number of segments used in the analysis. Figure 5 shows the correlation between TTPV-12SD or TTCO-12SD and Tmsv-16SD. Fair correlation was noted between two indices. Concordance rate between TTPV-12SD and Tmsv-16SD for detecting LV dyssynchrony was 79% (91/115). The corresponding value between TTCO-12SD and Tmsv-16SD was 80% (92/115). Because one of the advantages of RT3DE for the evaluation of LV dyssynchrony is the ability to assess the apex, we calculated SD of Tmsv in 4 apical regions (Tmsv-apexSD) in each patient. The mean value of Tmsv-apex was 38 ⫾ 50 milliseconds (range 0-312 milliseconds). Significant correlation was noted between Tmsv-apexSD and Tmsv-16SD (r ⫽ 0.80, P ⬍ .001). There was also significant correlation between Tmsv-apexSD and TTPV-12SD (r ⫽ 0.71, P ⬍ .001) and TTCO-12SD (r ⫽ 0.74, P ⬍ .001).

Measurement Variability The intraobserver variability of Tmsv-6SD, Tmsv12SD, and Tmsv-16SD by RT3DE was 29%, 26%, and 32%, respectively. The corresponding value of TTPV-12SD and TTCO-12SD by DTI was 20% and 16%, respectively. Interobserver variability of Tmsv6SD, Tmsv-12SD, and Tmsv-16SD by RT3DE was 26%, 39%, and 56%, respectively. The corresponding value of TTPV-12SD and TTCO-12SD by DTI was 41% and 22%, respectively.

DISCUSSION CRT is an effective nonpharmacologic treatment for heart failure associated with LV dysfunction.1-5 The noninvasive assessment of LV dyssynchrony has become increasingly important for selecting patients who will benefit from CRT. Recent data have demonstrated that mechanical dyssynchrony is not completely described by electrical dyssynchrony.7 Indeed, some patients with wide QRS complexes do not exhibit LV dyssynchrony, whereas some patients with a narrow QRS complex may demonstrate LV dyssynchrony.7,8 These considerations suggest that

1326 Takeuchi et al

the surface ECG may not be the optimal marker to select candidates for CRT. New imaging techniques, such as DTI, may be useful to select CRT responders. Previous studies have reported that measuring the time interval between the R wave of the ECG and peak systolic longitudinal velocity from myocardial

Journal of the American Society of Echocardiography December 2007

velocity curves obtained with DTI is useful to quantitate systolic asynchrony.5,10-12,14,15 In particular, the SD of the TTPV in the basal 6 and middle 6 LV segments by DTI has been shown to be a good predictor of LV reverse remodeling after CRT.5,15 Although DTI data are derived from high frame rate imaging that may provide more accurate intramyocardial velocity information, this index is derived from only 12 segments obtained from the 3 apical long-axis views obtained separately. This technique does not reflect dyssynchrony of the entire LV, because it is unable to obtain information from the apical region, as a result of the angle dependency of this technique. Finally, difficulties in positioning the ROI in the same place in serial studies may result in intermeasurement variability. RT3DE had been demonstrated to assess LV remodeling but additionally appears to be a potential alternative for the evaluation of LV dyssynchrony. By using the regional volume versus time changes combined with a fast semiautomated LV contour tracking algorithm, RT3DE acquired from 4 consecutive beats provides an LV dyssynchrony index that reflects the spatial motion of the entire ventricle. The apical cap is excluded from the analysis as it contributes little to ejection, unlike DTI, all other apical segments are analyzed. This results in a more complete assessment of apical region. Thus, LV remodeling (ie, LV volumes) and synchronicity can be assessed in one analysis. Current Study This is the first study to determine the feasibility and accuracy of measuring the LV dyssynchrony index using full-volume RT3DE data sets combined with a fast contour tracking algorithm. This technique requires manual determination of 5 specific points followed by automated contour tracking of the entire LV endocardium. The 5 points use 3D texture tracking to compensate for LV motion. The computer algorithm automatically defines and traces endocardial border in subsequent frames throughout the cardiac cycle without the need for tedious

Figure 3 Mean ⫾ SD and individual data plot of Tmsv6SD (upper panel), Tmsv-12SD (middle panel) and Tmsv16SD (lower panel) according to the degree of left ventricular ejection fraction. Dotted line, means mean ⫹ 2SD derived from healthy volunteer (n ⫽ 21). Parenthesis means number of subjects whose data exceeds dotted line. LVEF; Left ventricular ejection fraction, Tmsv-6SD; the SD of the time interval between the R wave and minimal end-systolic volume in 6 basal segments, Tmsv-12SD; the SD of the time interval between the R wave and minimal end-systolic volume in 6 basal and 6 middle segments, Tmsv-16SD; the SD of the time interval between the R wave and minimal end-systolic volume in 6 basal, 6 middle and 4 apical segments.

Journal of the American Society of Echocardiography Volume 20 Number 12

Takeuchi et al 1327

Figure 4 Mean ⫾ SD and individual data plot of TTPV-12SD (left) and TTCO-12SD (right) according to degree of left ventricular ejection fraction. Dotted line means mean ⫹ 2SD derived from healthy volunteer (n ⫽ 21). Parenthesis means number of subjects whose data exceeds dotted line. TTPV-12SD; The SD of time from the R wave to the peak systolic velocity in 6 basal and 6 middle segments, TTCO-12SD; the SD of time from the R wave to the cross over point in the temporal-axis in 6 basal and 6 middle segments. Table 2 Correlation between parameters of left ventricular dyssynchrony by real-time 3-dimensional echocardiography and Doppler tissue imaging TTPV-12SD (DTI)

TTCO-12SD (DTI)

RT3DE

r

P

r

P

Tmsv-16SD Tmsv-12SD Tmsv-6SD Tmsv-S-L Tmsv-S-PT

0.71 0.62 0.49 ⫺0.18 ⫺0.07

⬍.001 ⬍.001 ⬍.001 ns ns

0.73 0.66 0.49 ⫺0.20 ⫺0.08

⬍.001 ⬍.001 ⬍.001 ⬍.05 ns

DTI, Doppler tissue imaging; ns, not significant; RT3DE, real-time 3-dimensional echocardiography; Tmsv, time to minimal systolic regional volume; TTCO, time to crossover point in the x-axis; TTPV, time to peak systolic velocity.

manual tracing. The entire process averages about 60 seconds. Our results showed that both the intraobserver and interobserver variability of RT3DEderived dyssynchrony indices is larger than that recently published by Kapetanakis et al,16 who reported an intraobserver and interobserver variability for Tmsv-16SD of 6.4% and 8.1%, respectively, using other quantitative software. Reasons for this include: differences in the placement of the points required for the initialization of the automated border detection. As the algorithms for border detection improve, this variability should decrease. Our results also showed relatively large intraobserver and interobserver variability for DTI. With DTI, variability originates when the tissue profile contains more than two peaks, a situation that is frequently encountered in patients with severe LV dysfunction and wide QRS morphology. Also, differences in the placement of the ROI between observers can account for part of the variability.

Theoretically, RT3DE-derived Tmsv-16SD could represent the degree of mechanical dyssynchrony of the entire LV more completely than Tmsv-6SD and Tmsv-12SD. We observed a better correlation between TTPV-12SD and Tmsv-16SD than between TTPV-12SD and Tmsv-12(6)SD in this study. Because tissue Doppler velocity in the basal and middle part of the LV apparently contains information on apical motion, a better correlation between TTPV-12SD and Tmsv-16SD was predictable. Thus, Tmsv-16SD might be a novel and potentially useful index for the direct measurement of LV dyssynchrony. Previous studies have reported contradictory results about the correlation between 3D-derived LV dyssynchrony index and DTI-derived LV dyssynchrony index. Kapetanakis et al16 reported poor correlation between Tmsv-16SD and TTPV-12SD (r ⫽ 0.26, P ⫽ .06). Zhang et al17 showed fair correlation (r ⫽ 0.74, P ⬍ .05), which was in agreement of our results. Although concordance rate between two indices for detecting LV dyssynchrony in this study was good (80%), further studies are required as to which is a better predictor for the responder after CRT in a larger number of patients. We observed that Tmsv-16SD increased significantly with the severity of LV systolic dysfunction. When a cut-off value of Tmsv-16SD of greater than 36 milliseconds (mean ⫹2SD for healthy volunteers) was used to define significant mechanical dyssynchrony, significant LV dyssynchrony occurred in 67% of patients with LVEF less than 50% in contrast to 6% of patients with LVEF 50% or greater. These results indicate that the degree of LV dyssynchrony adversely affects global LV systolic function. Interestingly, one third of patients with LVEF ⬍ 50% who showed no LV dyssynchrony assessed by RT3DE in this study. This value is identical to the previously

1328 Takeuchi et al

Journal of the American Society of Echocardiography December 2007

Figure 5 Correlation between Tmsv-16SD and TTCO-12SD (left) and between Tmsv-16SD and TTPV-12SD (right). Dotted line cut-off value for each method. Tmsv-16SD; the SD of the time interval between the R wave and minimal end-systolic volume in 6 basal, 6 middle and 4 apical segments by real time 3D echocardiography. TTCO-12SD; The SD of time for the R wave to the cross over point in the temporal-axis in 6 basal and 6 middle segments by Doppler tissue imaging, TTPV-12SD; the SD of time from the R wave to the peak systolic velocity in 6 basal and 6 middle segments by Doppler tissue imaging.

reported number of nonresponder after CRT therapy. Finally, a poor correlation was noted between the time difference of Tmsv from only two LV segments and DTI-derived dyssynchrony index. It is likely that assessment of LV dyssynchrony in only two segments is not sufficient to evaluate the entire LV dyssynchrony.17 Limitations In this study, we did not perform ROI tracking for DTI. It could adversely affect the accuracy for absolute value of the myocardial velocity, but the profile of the velocity curve would be less affected. Placing the large size of ROI, which covered the entire myocardium in the specific segments, may also reduce this limitation. In contrast to DTI, the lower frame rate of RT3DE leads to inferior temporal resolution and, thus, cannot track subtle degrees of LV dyssynchrony between segments. Although in theory this may lead to underestimation of LV dyssynchrony, this minimal LV dyssynchrony between segments does not contribute much to regional volumetric ejection. Furthermore, the subtle assessment is not necessary for selecting patients with obvious LV dyssynchrony and, thus, the inferior temporal resolution is likely of minimal clinical significance. Image quality of the data set may also affect the measurements. Further refinement of the transducer technology will overcome these problems. Conclusions Although intraobserver and interobserver variability was somewhat large, the assessment of LV dyssynchrony by RT3DE and newly developed software is relatively easy. In addition to providing accurate

global LV parameters like volume and EF, good correlation between this index and DTI-derived LV dyssynchrony indices makes RT3DE useful modality for assessing LV dyssynchrony. REFERENCES 1. Abraham W, Fisher W, Smith A, Delurgio D, Leon A, Loh E, et al. Cardiac resynchronization in chronic heart failure. N Engl J Med 2002;346:1845-53. 2. Gras D, Leclercq C, Tang A, Bucknall C, Luttikhuis H, Kirstein-Pedersen A. Cardiac resynchronization therapy in advanced heart failure the multicenter InSync clinical study. Eur J Heart Fail 2002;4:311-20. 3. Leclercq C, Kass D. Retiming the failing heart: principles and current clinical status of cardiac resynchronization. J Am Coll Cardiol 2002;39:194-201. 4. Saxon L, De Marco T, Schafer J, Chatterjee K, Kumar U, Foster E, et al. Effects of long-term biventricular stimulation for resynchronization on echocardiographic measures of remodeling. Circulation 2002;105:1304-10. 5. Yu C, Chau E, Sanderson J, Fan K, Tang M, Fung W, et al. Tissue Doppler echocardiographic evidence of reverse remodeling and improved synchronicity by simultaneously delaying regional contraction after biventricular pacing therapy in heart failure. Circulation 2002;105:438-45. 6. Pitzalis M, Iacoviello M, Romito R, Massari F, Rizzon B, Luzzi G, et al. Cardiac resynchronization therapy tailored by echocardiographic evaluation of ventricular asynchrony. J Am Coll Cardiol 2002;40:1615-22. 7. Yu C, Fung W, Lin H, Zhang Q, Sanderson J, Lau C. Predictors of left ventricular reverse remodeling after cardiac resynchronization therapy for heart failure secondary to idiopathic dilated or ischemic cardiomyopathy. Am J Cardiol 2003;91:684-8. 8. Yu C, Lin H, Zhang Q, Sanderson J. High prevalence of left ventricular systolic and diastolic asynchrony in patients with congestive heart failure and normal QRS duration. Heart 2003;89:54-60.

Journal of the American Society of Echocardiography Volume 20 Number 12

9. Yu CM, Yang H, Lau CP, Wang Q, Wang S, Lam L, et al. Regional left ventricle mechanical asynchrony in patients with heart disease and normal QRS duration: implication for biventricular pacing therapy. Pacing Clin Electrophysiol 2003;26: 562-70. 10. Ansalone G, Giannantoni P, Ricci R, Trambaiolo P, Laurenti A, Fedele F, et al. Doppler myocardial imaging in patients with heart failure receiving biventricular pacing treatment. Am Heart J 2001;142:881-96. 11. Bax J, Marwick T, Molhoek S, Bleeker G, van Erven L, Boersma E, et al. Left ventricular dyssynchrony predicts benefit of cardiac resynchronization therapy in patients with endstage heart failure before pacemaker implantation. Am J Cardiol 2003;92:1238-40. 12. Bax J, Bleeker G, Marwick T, Molhoek S, Boersma E, Steendijk P, et al. Left ventricular dyssynchrony predicts response and prognosis after cardiac resynchronization therapy. J Am Coll Cardiol 2004;44:1834-40. 13. Garrigue S, Jaos P, Espil G, Labeque J, Hocini M, Shah D, et al. Comparison of chronic biventricular pacing between epicardial and endocardial left ventricular stimulation using Doppler tissue imaging in patients with heart failure. Am J Cardiol 2001;88:858-62. 14. Sogaard P, Egeblad H, Kim W, Jensen H, Pedersen A, Kristensen BO, et al. Tissue Doppler imaging predicts improved systolic performance and reversed left ventricular remodeling during long-term cardiac resynchronization therapy. J Am Coll Cardiol 2002;40:723-30. 15. Yu C, Fung J, Zhang Q, Chan C, Chan Y, Lin H, et al. Tissue Doppler imaging is superior to strain rate imaging and postsystolic shortening on the prediction of reverse remodeling in both ischemic and nonischemic heart failure after cardiac resynchronization therapy. Circulation 2004;110:66-73.

Takeuchi et al 1329

16. Kapetanakis S, Kearney M, Siva A, Gall N, Cooklin M, Monaghan M. Real-time three-dimensional echocardiography: a novel technique to quantify global left ventricular mechanical dyssynchrony. Circulation 2005;112:992-1000. 17. Zhang Q, Yu C, Fung J, Zhang Y, Chan Y, Chan H, et al. Assessment of the effect of cardiac resynchronization therapy on intraventricular mechanical synchronicity by regional volumetric changes. Am J Cardiol 2005;95:126-9. 18. Caiani E, Corsi C, Zamorano J, Sugeng L, MacEneaney P, Weinert L, et al. Improved semiautomated quantification of left ventricular volumes and ejection fraction using 3-dimensional echocardiography with a full matrix-array transducer: comparison with magnetic resonance imaging. J Am Soc Echocardiogr 2005;18:779-88. 19. Corsi C, Lang R, Veronesi F, Weinert L, Caiani E, MacEneaney P, et al. Volumetric quantification of global and regional left ventricular function from real-time three-dimensional echocardiographic images. Circulation 2005; 112:1161-70. 20. Jacobs LD, Salgo IS, Goonewardena S, Weinert L, Coon P, Bardo D, et al. Rapid online quantification of left ventricular volume from real-time three-dimensional echocardiographic data. Eur Heart J 2006;27:460-8. 21. Mor-Avi V, Sugeng L, Weinert L, MacEneaney P, Caiani E, Koch R, et al. Fast measurement of left ventricular mass with real-time three-dimensional echocardiography: comparison with magnetic resonance imaging. Circulation 2004;110: 1814-8. 22. Gerald O, Bilson A, JM. R, Jacob M, Fradkin M, Allouche C. Efficient model-based quantification of left ventricular function in 3-D echocardiography. IEEE Trans Med Imaging 2002;21:1059-68.