Introduction
Left ventricular (LV) strain by two-dimensional (2D) speckle-tracking echocardiography has been shown to be a reliable and clinically important method of quantitatively characterizing cardiac function in children [
1‐
3]. In addition, LV strain is correlated with synchrony [
3‐
6] and their interdependence plays an important role in managing patients with both congenital and acquired heart disease [
7‐
12]. Strain and synchrony can be calculated with the use of a variety of analysis packages, some of which are vendor-specific, while others are vendor-independent [
13]. Recent studies in adults [
14] and children [
15,
16], as well as an international task force for deformation imaging, have investigated the variations in reported values of strain from different packages [
10,
13]. A paucity of evidence on inter-vendor synchrony differences currently exists, with only a single study published in adults, and none in children [
4]. We hypothesized that different modern analysis packages may produce different results in both strain and synchrony values in children, independent of acquisition. Accordingly, the aim of this study was to evaluate the variability of analyses derived from speckle tracking echocardiography using three popular software platforms (QLAB 10.5, 10.8, and TomTec 1.2) for the measurement of longitudinal strain, circumferential strain, and synchrony.
Discussion
In this study we compared longitudinal and circumferential strain and synchrony measures using two commercially available and European Association of Cardiovascular Imaging/American Society of Echocardiography task force tested analysis tools (QLAB and TomTec) with several different settings. The main findings are the following: 1) the tracking feasibility and reproducibility of all measures with QLAB 10.5 differed significantly compared to QLAB 10.8 and TomTec packages (the latter two products fully incorporated the task force recommendations, while the former did not); 2) mean measured time to peak standard deviation (synchrony) differed among all vendors and settings; 3) tracking assessment and strain measures were reproducible and similar among the three smoothness settings for QLAB 10.8; 4) AP4 mean longitudinal and SAXM circumferential strain had minimal differences with good tracking assessment for TomTec at both high and low frame rates in healthy controls and ventricular paced hearts, but showed wider mean differences in patients with flattened interventricular septum at low frame rates. To our knowledge, this is the first study investigating the agreement between commonly used QLAB and TomTec platforms in pediatrics with different smoothness and frame rate settings. The findings in this study imply that the technical characteristics of the tested software packages coupled with the patient population likely have a combined impact on the measurement results.
Few studies have addressed the inter-vendor variability of longitudinal and circumferential strain measurements in children [
12,
15,
16,
18,
19]. The EACVI/ASE developed a Task Force to standardize deformation imaging and initially concluded from head-to-head comparison of nine different vendors in adults only (including TomTec Image Arena 2D CPA 1.2 and QLAB version 10.0) that global LV longitudinal strain was feasible and comparable, and in many cases superior to conventional echocardiographic parameters such as volumes and ejection fraction [
13]. Prior to the Task Force establishment, Koopman et al. showed that longitudinal strain, measured using a vendor-independent software (TomTec version 1.0) and a vendor-specific software (EchoPAC version 7.0 and QLAB version 7.0) had reasonable agreement, while for circumferential strain wide variability existed between the software packages [
15]. Ramlogan et al. recently assessed the measured variability of longitudinal and circumferential strain in children with normal cardiac segmental anatomy and varying cardiac function between two post-EACVI/ASE Task Force recommended software platforms, TomTec (Image-Arena ver. 4.6 SP3 CPA 1.2) and GE (EchoPAC version BT13); and demonstrated reasonable comparison between vendors, with overall longitudinal strain agreement slightly stronger than circumferential [
16]. de Waal et al. also reported robust longitudinal strain agreement in term and preterm infants between QLAB 10.8 and TomTec CPA 1.3, but poor correlation with circumferential strain [
18]. The data in our study are in line with previous reports in children with respect to similarities in longitudinal strain among vendors, but also suggest a fairly robust agreement with circumferential and longitudinal strain between QLAB 10.8 and TomTec 1.2 in children.
The factors that modulate variability in strain measurement by speckle tracking echocardiography include (1) variability in image acquisition, (2) intra and inter- observer differences in postprocessing, and (3) differences between proprietary software for image analysis; moreover, these differences may be impacted by differences in the pattern of ventricular contraction [
11,
20,
21]. Since we analyzed longitudinal and circumferential strain and synchrony from the same AP4 and SAXM image, respectively, for each vendor package and setting, respectively, our design had no bias that may be incurred from use of different images. In order to determine the tracking capability of each software package we utilized a previously validated image-grading system defined by our laboratory [
11]. The ability to manually adjust the endocardial border in end-systole in QLAB 10.8 (all three smoothness levels) and TomTec (low and high frame rate), likely resulted in their higher tracking scores, compared to QLAB 10.5. To address postprocessing and software variability, TomTec offers the possibility to perform off-line deformation analysis on archived images acquired by any platform, thus providing a clinically useful tool to analyze and compare studies performed in different centers with different echocardiography platforms. The images imported into TomTec are downsampled to low frame rates, influencing temporal resolution and possible hindering reliable assessment of both peak strain values and timing indices. Low frame rates have also been hypothesized to decrease the agreement of strain and synchrony measurements with other vendors that analyze the images at high frame rates. TomTec also imports images with lower fidelity compared to the raw image files used by the vendor dependent platforms, potentially influencing spatial resolution and decreasing the tracking reliability of the speckle patterns [
16]. In our study, we demonstrated that LV longitudinal and circumferential strain showed excellent agreement when using TomTec at high and low frame rates in children with normal cardiac anatomy and ventricular paced rhythm, similar to recent reports in children [
15,
16], and adults [
17]. For children with abnormal septal contour, however, differences in mean longitudinal and circumferential strain between the high and low frame rate analysis were larger. Given that ventricular volume loading status of the patient does affect the analysis [
21], our study demonstrates that it may be rational to obtain both longitudinal and circumferential strain in archived images (low frame rates), but it is inconsistent with the previously suggested notion that archived, low temporal resolution images are adequate for all analyses [
21].
Mean strain from all three smoothness settings from QLAB 10.8 were similar. Version 10.7 of the QLAB software was the first to employ the different smoothness settings. A user has the option to preset the quantification (low, medium, or high values) that predefines the assumed smoothness of tissue motion used for the speckle-tracking algorithm by putting the width of a kernel around the tracking points of interest. A higher smoothness quantification setting signifies larger kernels around the region of interest, resulting in less sensitivity of the algorithm to local tissue motion. Our study demonstrated minimal differences in longitudinal and circumferential strain values among the low, medium and high settings for QLAB 10.8, but wide mean differences for TTPSD values. Although we previously demonstrated good reproducibility (intra- and inter- observer reliability) of TTPSD values with QLAB 10.5 at a high frame rate and a single smoothness setting [
3], the wider variability of TTPSD using the three different smoothness settings for QLAB 10.8 is not surprising. Whereas the location of the peak of the global strain curve can be adjusted in systole, the boundaries of the constituent segmental curves (determining the TTPSD) cannot be modified accordingly.
Similar to strain, synchrony has been associated with clinical outcomes in children with congenital and acquired heart diseases [
7,
8], and a recent study by our echocardiography laboratory identified normal synchrony patterns and z-scores in children [
3]. Since ventricular synchrony measures are associated with life-threatening arrhythmias, heart failure, and mortality in pediatric populations [
7,
8], understanding the variability between vendors might allow the generalizability of these findings across echocardiography laboratories. Van Everdingen et al. demonstrated minimal differences in longitudinal TTPSD with adults (patients had a range of structurally normal hearts and varying degrees of cardiac function and arrhythmias) between QLAB (10.0) and TomTec (1.2.1.2) [
4]. Values were lower (more synchronous) in QLAB compared to the TomTec. In our work, we also observed lower AP4 and SAXM TTPSD means in the QLAB 10.5 and 10.8 (high smoothness setting) compared to TomTec. Although the mechanism for these differences are not completely clear, it is likely that the suboptimal tracking throughout the cardiac cycle (in QLAB 10.5) and spatial averaging effects (in QAB 10.8, high smoothness setting) contribute to an underestimation of dyssynchrony. Overall, we found very wide variation in measures between all the software packages and settings for longitudinal and circumferential TTPSD. Although TTPSD may be reproducible amongst the same vendors in adults [
4,
22] and children [
3], our data suggest avoidance of interchanging the TTPSD measures between studies using different vendor software packages and settings [
3].
Limitations
The importance of these results should be interpreted within the framework of the inherent limitations in this study. First, while the cardiac cycles analyzed using the different software packages always came from the same acquired clips (typically 2–3 beats in duration), they were not necessarily from the same beat (the beat selected for analysis was the one that was best tracked by the software analysis package); however, given the temporal proximity of all of the beats in the same clip, they would all be expected to reflect a very similar hemodynamic state. Secondly, our study did not analyze the variability with comparison of serial measurements between vendors. Currently, the EACVI/ASE Task Force and other independent studies in infants, children and adults suggest that serial measurements might not be interchangeable when conducting a longitudinal study [
15,
18,
19] and “should be preferably interpreted relative to previous examinations with the same machine and software” [
13]. In practice “Global” LV strain has been defined by different LV segmentation models that assesses either an 18-, 17- or 16- segment model from the averaging of the three apical views (for longitudinal strain) and the three short axis views (for circumferential strain) [
10,
13,
23]. In children deformation imaging is often reported only from the AP4 view for longitudinal strain and SAXM for circumferential [
1,
2,
13] because it is sometimes difficult to show an apex clearly as it exists near the surface and it is not always contained in the view [
24]. Moreover, not every package that can be used in pediatrics was included in this study; however, the products chosen represent some of the most popular and commonly used tools available. Finally, no attempt was made to directly compare the values or reproducibility of regional strain measurements; however the analysis of ventricular synchrony that was performed in this work does address issues of regional functional assessment.
Conclusions
This study investigated the agreement between commonly used QLAB and TomTec platforms in pediatrics with different smoothness and temporal resolution settings. Mean synchrony (TTPSD) varied substantially among all QLAB and TomTec packages and settings in children. Mean longitudinal (AP4) and circumferential (SAMX) strain also differed between QLAB 10.5 and all other software packages, irrespective of the temporal or spatial resolutions, suggesting avoidance of QLAB 10.5 comparisons with contemporary QLAB versions and vendor-independent TomTec software. For QLAB 10.8, the various quantification smoothness settings resulted in minimal variability in SAXM and AP4 views. Compared with QLAB 10.8, there were also minimal differences with TomTec for both longitudinal and circumferential strain values, even with the high and low frame rate settings, for the normal and ventricular paced cohorts; for children with abnormal septal contours, mean differences according to package/setting were greater. These findings suggest that reliable comparisons among these tools, using different settings, may be appropriate in many settings, but clinicians and researchers should use these tools with caution.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.