Speckle-tracking strain is almost universally cited as being independent of angle of insonation, but there are minimal confirmatory studies, and this claim may not be consistent with the known limitations of ultrasound axial and lateral spatial resolution. The aim of this study was to assess the influence of angle and depth on longitudinal peak systolic strain (LPS).
Thirty-four healthy pediatric subjects (age range, 6–18 years; 47% male) with normal cardiac anatomy and good image quality were prospectively imaged. Angular comparisons of LPS were investigated by examining interangle reproducibility on the basis of one standard and one alternative image acquisition relative to intraobserver reproducibility of two standard views of the same left ventricular segments. A single-window comparison was used to evaluated septal LPS: standard apical four-chamber versus right ventricular centered four-chamber. Two paired standard and alternative window comparisons were as follows: (1) four-chamber: standard apical versus subcostal; and (2) three-chamber: standard apical versus parasternal long-axis.
The global LPS intraobserver difference using the paired standard and alternative window comparisons was lower than the interangle difference in global LPS (−1.0 ± 0.1% vs −2.1 ± 2.4%). Intraobserver reproducibility was significantly higher than interangle reproducibility (intraclass correlation coefficient = 0.9 vs 0.29, P < .001). Similar results were found in the segmental strain analysis. Interangle reproducibility was significantly decreased compared with intraobserver reproducibility in the septal single-window comparison. Target depth assessment demonstrated a systematic bias between the near-field and far-field segments.
Echocardiographically derived LPS values were modestly dependent on angle of insonation and target depth in this pediatric population. Normal strain ranges derived from standard apical images should not be applied to strain derived from sub-costal images, off-axis apical imaging, or applications in which a standard window cannot be defined.
Speckle-tracking echocardiographic (STE) strain analysis is almost universally cited as being independent of angle of insonation and target depth, but this may not be consistent with known limitations of ultrasound imaging and has not been confirmed in clinical studies. Speckle-tracking echocardiography is increasingly used in clinical and research settings to provide functional and dyssynchrony assessments in congenital and acquired heart disease. Frequently, regional contraction timing is used to assess mechanical dyssynchrony to predict response to cardiac resynchronization therapy in patients with left bundle branch block or pacemaker-induced cardiomyopathy, and global strain is used as a quantitative marker of function.
The STE method tracks speckles that are natural acoustic echodensities (backscatter) that result from the complex myofiber arrangement within the B-mode image. Speckle-tracking echocardiography is frequently considered to be superior to the previous tissue Doppler method, which is highly dependent on the angle of insonation. Longitudinal peak systolic strain (LPS) is a reproducible assessment of ventricular mechanics, and although the basic calculation of strain is angle independent, the use of B-mode images for strain calculation may alter that assumption. B-mode echocardiographic imaging is fundamentally anisotropic, with differences between lateral and axial resolution and dependence on the wavelength, transducer aperture, frame rate, line sequencing, focal zone, target depth, and angle of insonation. These factors may create a degree of angular dependence.
Early STE studies in open-chested sheep comparing sonomicrometry with LPS demonstrated good agreement (intraclass correlation coefficient [ICC] = 0.8) between strain values. Clinical studies testing the effect of angle and target depth on LPS are lacking. The goal of this study was to test the assumption that strain is angle and depth independent by comparing LPS obtained from paired views of the same left ventricular (LV) segments from different transducer angles and target depths in healthy children. In this study, angle was defined between the imaging scan line and the myocardial wall of interest.
In this cohort study, we prospectively obtained echocardiograms in 34 pediatric subjects with normal cardiac anatomy and good acoustic windows. Echocardiography was clinically ordered to assess ventricular function in the setting of a rheumatologic or oncologic disease while on chemotherapeutic agents. All study examinations included three paired views optimized for STE. Exclusion criteria were any congenital cardiac anatomic abnormalities, poor acoustic windows, pacemakers, and history of dysrhythmias. Studies were performed between June 2012 and February 2013. Institutional review board approval and informed consent were obtained from all subjects.
Echocardiography and Strain Analysis with Three Paired Views
All echocardiographic studies were acquired on a GE Vivid 7 using a 5-MHz ultrasound probe (GE Vingmed Ultrasound AS, Horten, Norway). The echocardiographic examination was performed with grayscale images optimized for strain analysis (70–100 frames/sec) with the three paired views and subjects in the standard left lateral recumbent position. Imaging depth and focal zone placement were adjusted for optimum image quality. Sonographers received specialized STE strain analysis training to ensure proper optimization of frame rates, view acquisition, and image quality. Two of the three paired views were obtained from paired standard and alternative windows (PSAWs) to evaluate differences in LPS: (1) apical three-chamber versus parasternal (alternative PLAX3) and (2) apical four-chamber (standard LV4) versus subcostal (alternative SC4) ( Figure 1 ). These windows were chosen to provide near-orthogonal evaluations of the same ventricular walls. The apical segments in the three-chamber comparison were excluded because of difficulty acquiring the apical region within the sector arc on the PLAX3 view.
The third paired view compared middle and basal septal LPS from a single-window comparison between the four-chamber LV apical view (standard LV4) and a four-chamber apical view centered on the right ventricle (alternative RV4) to compare the importance of minor angular differences in axial alignment obtained from the same imaging window ( Figure 1 ). The septum was at modestly different angles to the scan lines in those two views (LV4 and RV4). The angle between the septal plane and the scan line intersecting the midpoint of the septum was calculated by tracing the septum and scan line from the screen to paper and determining the angle manually using a protractor.
Offline strain analysis was performed by using EchoPac PC version BT11 (GE Vingmed Ultrasound AS) by an experienced single reader (D.F.). The reference point was placed at QRS onset. The aortic valve closure timing relative to the QRS complex was defined using spectral Doppler from the apical view. The endocardial border was traced in end-systole for speckle-tracking. The region of interest was adjusted to exclude the pericardium, papillary muscles, and chordal apparatus. The integrity of speckle-tracking was automatically detected and visually confirmed. In case of poor tracking, tracing was readjusted or the shape of the region of interest changed. Longitudinal segments were divided by the software into basal, middle, and apical segments on the septal and lateral walls. LPS was calculated in six segments during systole between QRS onset and aortic valve closure in the PSAW comparisons. In the single-window comparison, septal LPS from the LV4 and RV4 views was obtained by tracing the septum with software segmentation into three segments. The septum began to diverge into the left and right ventricles apically, so only middle and basal segmental LPS was included for analysis. This software has been validated for the determination of LPS in the left ventricle.
Unless otherwise stated, continuous variables are reported as mean LPS ± SD. For the three paired measurements, the difference in LPS between views was calculated to produce descriptive summary statistics of intraobserver and interangle reproducibility. The interangle difference was defined as the LPS difference between the standard and alternative views. The intraobserver difference was defined as the LPS difference between repeated analyses of the standard view by a single reader (intraobserver difference) to evaluate within-reader variability inherent in strain analysis. For the intraobserver difference, repeated strain analysis was performed de novo with a new region of interest traced on the same cardiac cycle for all standard views by the same reader ≥3 weeks after the first read, blinded to previous results. Paired t tests were used to determine whether these differences were statistically significant. The Spearman rank correlation was also computed to assess whether there were correlation between the paired measurements.
If strain is angle dependent, one would expect the interangle reproducibility to be worse than the intraobserver variability, so that strain based on the alternative (angled) view cannot be used interchangeably with strain based on the standard view. To claim that there is angle dependence in strain, the null hypothesis is that the intraobserver reproducibility is the same as the interangle reproducibility, where reproducibility is assessed by ICC.
The test of this hypothesis was carried out by using the Z -score test statistic with the estimated difference divided by the standard error estimated on the basis of bootstrap samples, where the bootstrap sampling unit was all three paired measurements together to account for correlation due to repeated measures on the same subject or segment. Specifically, 1,000 bootstrap samples were taken and the corresponding 1,000 intraobserver ICCs, interangle ICCs, and differences of intraobserver ICCs and interangle ICCs were calculated. The Z score was computed as the observed difference of intraobserver ICC and inter-angle ICC divided by the standard deviation of the 1,000 bootstrap differences of intraobserver and interangle ICCs. The P value of the hypothesis test was obtained by comparing the Z score with the standard normal distribution.
For the PSAW comparisons, the same basal and middle segments from the four- and three-chamber view comparisons were analyzed together because of similarities in angle changes between views and target position within the sector arc (i.e., basal septal segmental LPS from the four- and three-chamber view comparisons were grouped for a total of 34 × 2 = 68 segments). Apical segmental LPS could not be obtained in the PLAX3 view, and therefore only four-chamber apical segment comparisons are reported.
To test for dependence on target depth, the average LPS mean difference in the middle and basal segments that are near field in the alternative view (septal) were compared with the averaged differences in the far-field alternative view segments (lateral wall).
P values < .05 were considered statistically significant. An a priori power analysis was performed to demonstrate that 32 subjects would have 80% power to detect a 50% increase in the variability (standard deviation) between the interangle and the intraobserver difference in the middle and basal segments at a level of .05. All statistical analyses were performed using Stata version 12 (StataCorp LP, College Station, TX) and SAS version 9.3 (SAS Institute Inc, Cary, NC).
This cohort consisted of 34 pediatric subjects with a mean age of 14.1 years (range, 6.7–17.9 years), of whom 47% were male. All subjects had normal cardiac anatomy and good parasternal, apical, and subcostal acoustic windows. Normal LV systolic function as assessed by global LPS > 16% was present in 33 of 34 subjects (97%). These subjects were all classified as having normal qualitative LV systolic function on the echocardiographic report. The remaining subject had global LPS of 14.4% with mildly diminished qualitative LV systolic function on a cardiotoxic chemotherapeutic agent. Adequate strain analysis tracking was present in 93% of the two standard views (378 of 408 total segments), and 93% of the two alternative views (315 of 340 total segments) excluding the PLAX3 apical segments. There were 374 paired four-chamber comparisons and 309 paired three-chamber comparisons.
Impact of Angle of Insonation
The descriptive statistics of the three paired measurements are reported in Table 1 , in which the global (averaged) values of LPS are listed in the PSAW comparison section. Global LPS was moderately correlated between standard and alternative views. The intraobserver reproducibility was strong compared with the interangle reproducibility, and these values were significantly different.
|View 1 standard||n ∗||View 1||View 1 repeated||P ∗||View 2 alternative||n †||View 2||P †||R ‡||P ‡|
|Single window comparison|
|Basal septal LPS||LV4||33||−17.1 ± 2.7%||−16.8 ± 3.1%||.15||RV4||33||−17.0 ± 2.4%||.76||0.44||.011|
|Midseptal LPS||LV4||33||−18.1 ± 2.7%||−17.8 ± 3.1%||.062||RV4||33||−17.9 ± 2.6%||.65||0.46||.007|
|Global LPS||34||−19.9 ± 2.1%||−20.0 ± 2.1%||.74||34||−22.0 ± 2.8%||<.001||0.56||<.001|
|Basal septal LPS||LV4/LV3||68||−18.8 ± 3.6%||−18.6 ± 4.0%||.29||SC4/PLAX3||63||−20.4 ± 5.1%||.025||0.45||.008|
|Midseptal LPS||LV4/LV3||68||−20.4 ± 3.2%||−20.3 ± 3.4%||.44||SC4/PLAX3||64||−22.1 ± 5.2%||.015||0.49||.033|
|Apical septal LPS||LV4||34||−23.1 ± 4.5%||−23.3 ± 4.4%||.63||SC4||25||−22.7 ± 4.6%||.24||0.12||.55|
|Apical lateral LPS||LV4||29||−21.4 ± 5.0%||−21.9 ± 4.8%||.38||SC4||22||−24.0 ± 4.4%||.12||−0.03||.87|
|Midlateral LPS||LV4/LV3||66||−18.8 ± 3.4%||−18.7 ± 3.4%||.50||SC4/PLAX3||62||−22.2 ± 4.6%||<.001||0.35||.042|
|Basal lateral LPS||LV4/LV3||59||−19.7 ± 4.4%||−19.8 ± 4.2%||.75||SC4/PLAX3||54||−22.9 ± 5.3%||<.001||0.41||.020|
Segmental LPS comparisons and correlations are also reported in Table 1 . In the single-window comparison, only middle and basal septal LPS was evaluated. For the PSAW comparisons, the segmental alternative view had significantly larger negative value than from the standard view, whereas there were no significant differences between the two standard views. The apical segments had higher variability and fewer samples because of their exclusion from the three-chamber comparison. No significant differences were seen between the two standard views or between the standard and alternative views.
Agreement between paired regional LPS comparisons is reported in Table 2 . The PSAW intraobserver difference in global LPS was lower than the interangle difference in global LPS (−1.0 ± 0.1% vs −2.1 ± 2.4%). Intraobserver reproducibility was significantly higher than interangle reproducibility (ICC = 0.9 vs 0.29, P < .001). Similar results were found in the segmental strain analysis. There was also significantly decreased interangle reproducibility compared with intraobserver reproducibility in the septal single-window comparison. Although the single-window comparison did not demonstrate systematic bias in the septal segments, the amount of variability was higher in the interangle comparisons, which resulted in a significantly lower interangle ICC than intraobserver ICC. These results are demonstrated visually in Bland-Altman plots of the intraobserver and interangle differences of global LPS in the PSAW comparison and basal septal LPS in the single-window comparison ( Figure 2 ). These plots provide representative examples from both the multiple- and single-window comparisons that were quite similar to the plots of the other conditions. In the single-window comparison, the angle between septal axis and the scan line averaged 9° (range, 1°–32°).