Two-dimensional STE is a cardiac imaging technique that allows noninvasive assessment of LV myocardial function with angle-independent deformation analysis.1 To perform deformation analysis, 2-D STE involves use of a postprocessing computer algorithm that tracks over time the motion of fixed patterns of speckles, which are natural artifacts generated by reflection, refraction, and scattering of ultrasonographic waves.1–6 Deformation variables assessed by use of 2-D STE include the so-called Lagrangian strain, which is defined as the comparison of the length of a myocardial segment at a given point in time, usually end-systole, with a reference length, usually its relaxed length at end-diastole, and SR, which represents the change in length per unit of time.2,4,6
Various vendor-dependent and -independent software systems have been developed for strain analysis by use of 2-D STE.7 Differences in derived results among these systems have been detected in human medicine, with postprocessing procedures an important determinant of variability in results among software plaftorms.8 An EACVI-ASE task force assessed and attempted to standardize GLS results. The task force identified moderate, significant variation in GLS values determined by various systems, with an absolute difference in strain units between systems of up to 3.7% in the same patient, which could be important when performing serial evaluations.8
Speckle tracking echocardiography has been validated against sonomicrometry in dogs,9 and various deformation variables have good correlation with invasive indices of LV function.10 Moreover, adequate feasibility and reproducibility for clinical and research purposes have been reported in healthy awake and sedated dogs.1,11–15 Several indices derived from 2-D STE have also been assessed in canine patients with myxomatous mitral valve disease of varying severity,16–19 dilated cardiomyopathy,1,15 and patent ductus arteriosus.1,11,20–22 Finally, 2-D STE has been used in several models of cardiac disease, including Duchenne muscular dystrophy.23 Published data on veterinary 2-D STE have mainly been for vendor-dependent STE software, including various versions of several plaftorms.1,10–23
To the authors' knowledge, the comparability of strain variables for dogs obtained by use of different software has not been investigated. Therefore, the objective of the study reported here was to compare 2-D STE-derived longitudinal deformation parameters obtained by use of vendor-specific and vendor-independent software for dogs with and without cardiac disease. The hypothesis was that results for the various software systems would have only moderate agreement for longitudinal global variables.
Materials and Methods
Animals
The study population comprised 2 groups of dogs. One cohort was from the Veterinary Teaching Hospital of the University of Murcia; images for this cohort were obtained with the ultrasound unita for vendor 1 (vendor 1 cohort). The other cohort was from the Hospital for Small Animals of the University of Edinburgh; images for this cohort were obtained with the ultrasound unitb for vendor 2 (vendor 2 cohort). For the vendor 1 cohort, cardiovascularly healthy dogs and client-owned dogs with cardiac disease of any age and body weight were prospectively enrolled between January 2015 and June 2017. Dogs with cardiac disease were recruited during cardiology consultations, whereas healthy dogs were owned by staff members and students. For the vendor 2 cohort, cardiovascularly healthy dogs and client-owned dogs with cardiac disease were recruited between October 2017 and January 2018. Client-owned dogs were identified during examination by the cardiopulmonary service.
Dogs were eligible for inclusion when they had undergone complete ECG-gated echocardiography that included good-quality images obtained at an adequate frame rate (between 60 and 100 Hz) to allow reliable longitudinal strain analysis. Dogs were considered cardiovascularly healthy or affected by cardiac disease on the basis of the medical history and results of a physical examination, CBC, serum biochemical analysis, noninvasive blood pressure measurement, ECG, conventional echocardiography, and, if needed, thoracic radiography. Owner consent was obtained for each dog before enrollment in the study. The study protocol was approved by the Ethics Committees of the Universities of Murcia and Edinburgh.
Echocardiographic examination
Transthoracic echocardiography at the University of Murcia was performed by a board-certified veterinary cardiologist (JFP) or a supervised resident in a cardiology training program (GS) with the ultrasound unit of vendor 1, which was equipped with a multifrequency 1- to 5-MHz phased-array sector transducer. Transthoracic echocardiography at the University of Edinburgh was performed by a board-eligible veterinary cardiologist (GS) or a supervised resident in a cardiology training program (JB) with the ultrasound unit of vendor 2, which was equipped with multifrequency 1.5- to 4.6-MHz and 2.4- to 8.0-MHz phased-array sector transducers.
Echocardiography was performed with simultaneous recording of the ECG and in accordance with published veterinary recommendations.24–26 Images recorded at a frame rate between 60 and 100 Hz for at least 4 consecutive cardiac cycles were used for 2-D STE analysis.
2-D STE analysis of echocardiographic images for the vendor 1 system
Echocardiographic examinations obtained with the ultrasound machine of vendor 1 were stored digitally for offline analysis and exported in standard formats to vendor 1-specific softwarec and as DICOM files to vendor-independent software.d Longitudinal strain and SR analyses were performed by a single operator (GS) with vendor 1-specific software and vendor-independent software. Analyses were performed with the same good-quality 2-D cine loop recorded from the left apical 4-chamber view.
For the vendor 1 software, longitudinal strain was obtained by manual placement of 3 reference points (one on each side of the mitral valve annulus and the third at the apical endocardial border). The software then created an ROI with the automatic delineation of endocardial and epicardial borders, and the operator adjusted the ROI to best incorporate the entire myocardial thickness. The LV was divided into 7 segments (basal septal, midseptal, apical septal, apical, apical lateral, midlateral, and basal lateral) by the software. The operator verified that the ROI followed the movements of each segment of the myocardium throughout the cardiac cycle; the operator performed manual adjustments when necessary. The software algorithm performed speckle tracking across the myocardium, and a weighted mean of the myocardial deformation (with the weighting greatest at the endocardium) was used for strain calculation. Derived segmental strain, global strain, and SR curves were generated, with the global waveform being created by use of a weighted mean of the regional strains. Analysis was performed by the vendor 1 software. If curves were not considered adequate and reliable, the dog was excluded from the study. Values for GLS, which were automatically provided by the software, were recorded; the values corresponded to the maximum strain or peak strain of the GLS waveform over the entire cardiac cycle. Peak systolic and diastolic SR values, which were displayed after placing the cursor at the appropriate time point along the global SR curves, were also recorded.
For the vendor-independent software, the endocardial border of the myocardium was manually traced from an end-diastolic frame starting at the septal side of the mitral valve annulus. The software automatically marked the epicardial border, and the operator verified that the ROI followed the movements of each segment of the myocardium throughout the cardiac cycle, making adjustments to the tracings when necessary. The software algorithm performed speckle tracking of 6 segments (basal septal, midseptal, apical septal, apical lateral, midlateral, and basal lateral) at the endocardial and epicardial regions; however, strain values for the epicardial region were excluded from analysis because the values sometimes were outside the echocardiographic window. Time-domain LV strain curves for each segment were created. The GLS indices were calculated as the mean for the peak segmental strain indices. The GLS values provided by the software (which corresponded to global endocardial peak longitudinal strains) and systolic endocardial SR values were recorded by the operator. Diastolic SR values, which were displayed after placing the cursor at the appropriate time point along the global endocardial SR curves, were also recorded.
2-D STE analysis of echocardiography images for the vendor 2 system
Echocardiographic examinations obtained with the ultrasound system for vendor 2 were stored digitally for offline analysis and exported in standard formats to vendor 2-specific softwaree and as DICOM files to the same vendor-independent software used for the vendor 1 cohort. Longitudinal strain and SR analyses were performed by a single operator (GS) with vendor 2-specific software and vendor-independent software. Analyses were performed with the same good-quality 2-D cine loop recorded from the left apical 4-chamber view.
For vendor 2 software, the endocardial border of the myocardium was manually traced from an end-systolic frame starting at the septal side of the mitral valve annulus, after which the software automatically created an ROI that the operator could adjust to best incorporate the entire myocardial thickness. The software algorithm then performed speckle tracking of 6 segments (basal septal, midseptal, apical septal, apical lateral, midlateral, and basal lateral) and assessed tracking quality. Segments marked as having inadequate tracking quality were retraced. If unreliable segments persisted after retracing was performed, the dog was excluded from the study. For segments with adequate tracking quality, segmental strain, GLS, and SR measured at the mid-myocardial layer were quantified by the software, and the global waveform was created by use of the mean of the regional strains. The GLS was automatically calculated by the software and displayed by default; it corresponded to the peak mid-myocardial longitudinal strain value, which was verified by the operator and then recorded. Peak systolic and diastolic SR values were recorded from the appropriate time point along the global SR curves. The GLS value obtained from the endocardial layer of the LV myocardium (which was displayed by the software as requested by the operator) was also recorded.
For the vendor-independent software, the same technique described for evaluation of echocardiographic images for vendor 1 was used.
Measurement reliability
Intraobserver within-day and between-day variability as well as interobserver variability were determined. Six echocardiograms obtained with each ultrasound system were arbitrarily selected, and the cine loops were subjected to repeated analyses by the same operator (GS) 2 times on the same day and on 2 separate days. For interobserver variability, a second investigator (MBT for images obtained with the ultrasound system for vendor 1 and JB for images obtained with the ultrasound system for vendor 2) performed separate repeated analyses on the same cine loops that the first operator used for assessment of intraobserver variability.
Images obtained with the ultrasound system for vendor 1 were assessed with vendor 1-specific software and vendor-independent software, and images obtained with the ultrasound system for vendor 2 were assessed with vendor 2-specific software and vendor-independent software. Observers were unaware of all other results.
Statistical analysis
Statistical analysis was performed with commercially available software.f Distribution of variables was tested for normality by use of the Shapiro-Wilk test (α = 0.05). Mean and SD were calculated for normally distributed variables, and median and range were calculated for nonnormally distributed variables. Study population characteristics were analyzed with descriptive statistics. For all analyses, values were considered significant at P < 0.05.
For the repeatability evaluation, mean and SD values resulting from vendor-dependent and vendor-independent strain analysis of the 6 repeated examinations obtained with either ultrasound system were used to determine interobserver and within-day and between-day intraobserver CVs for each software and were quantified as follows: CV = (SD of the measurements/mean of measurements) × 100. Values < 15% were considered adequate for clinical use.9
A paired t test or Wilcoxon signed rank test was used to make comparisons between the 2-D STE variables obtained with vendor-dependent and vendor-independent software. For each pair of values, the LOA and systematic errors were assessed by calculating the mean difference (bias) and the SD of the differences by use of Bland-Altman analysis.
Results
Study population
Ninety-six dogs were initially enrolled in the study (55 in the vendor 1 cohort and 41 in the vendor 2 cohort). Eleven dogs of the vendor 1 cohort and 1 dog of the vendor 2 cohort were subsequently excluded because of suboptimal tracking with the vendor-independent software.
The final vendor 1 cohort comprised 44 dogs (18 sexually intact males, 12 spayed females, 10 sexually intact females, and 4 castrated males). Age ranged from 1 to 17 years (median, 6.5 years), and body weight ranged from 3 to 48 kg (median, 21 kg). Fourteen dogs were mixed-breed dogs; the other 30 dogs were purebreds of 17 breeds, with the most common being Labrador Retriever (n = 4). Twenty-three dogs were cardiovascularly healthy, and 21 dogs had cardiac disease. Dogs with cardiac disease comprised 14 with myxomatous mitral valve disease (4 at stage B1, 5 at stage B2, and 5 at stage C, as determined on the basis of the American College of Veterinary Internal Medicine classification system27), 2 with preclinical idiopathic dilated cardiomyopathy, 2 with concurrent mild subaortic and pulmonic stenosis, and 1 each with mild subaortic stenosis, hypertensive LV hypertrophy, and right auricular hemangiosarcoma.
The final vendor 2 cohort comprised 40 dogs (17 spayed females, 13 castrated males, 9 sexually intact males, and 1 sexually intact female). Age ranged from 6 months to 14 years (median, 7 years), and body weight ranged from 1.3 to 47.5 kg (median, 21.2 kg). Four dogs were mixed-breed dogs; the other 36 dogs were purebreds of 22 breeds, with the most common being Labrador Retriever (n = 8) and German Shepherd Dog (4). Eighteen dogs were cardiovascularly healthy, and 22 dogs had cardiac disease. Dogs with cardiac disease comprised 8 with myxomatous mitral valve disease (5 at stage B1 and 3 at stage B2), 4 with preclinical idiopathic dilated cardiomyopathy, 3 with left-to-right shunting patent ductus arteriosus, 3 with severe pulmonic stenosis, 2 with tricuspid valve dysplasia, and 1 each with mild subaortic stenosis and pericardial hemangiosarcoma.
2-D STE
Results for 2-D STE variables for each software system (vendor 1, vendor 2, and vendor independent) were determined (Table 1), including GLS, as measured by use of each system (Figures 1 and 2).
Median (range) values for STE-derived longitudinal strain and SR obtained with vendor-dependent and vendor-independent software systems for cardiovascularly healthy dogs and dogs with various cardiac diseases.
Vendor 1 cohort (n = 44) | Vendor 2 cohort (n = 40) | |||||
---|---|---|---|---|---|---|
Variable | Vendor 1 | Independent | P value | Vendor 2 | Independent | P value |
GLS (%)* | ||||||
Transmural† to endocardial | −21.0 (−33.0 to −14.0) | −19.5 (−39.0 to −11.0) | 0.020 | — | — | — |
Mid-myocardial to endocardial | — | — | — | −17.5 (−27.5 to −9.5) | −18.8 (−28.9 to −1.3) | 0.006 |
Endocardial to endocardial | — | — | — | −21.7 (−32.3 to −11.4) | −18.8 (−28.9 to −1.3) | < 0.001 |
Systolic SR (1/s) | −2.2 (−3.2 to −1.4) | −1.8 (−4.4 to −0.7) | 0.005 | −1.8 (−2.9 to −0.3) | −1.6 (−2.9 to −0.4) | < 0.001 |
Early diastolic SR (1/s) | 2.1 (0.1 to 3.4) | 2.0 (0.6 to 5.4) | 0.357 | 1.6 (0.5 to 5) | 1.2 (0.4 to 4.0) | < 0.001 |
Late diastolic SR (1/s) | 1.5 (0.7 to 3.5) | 0.9 (0.3 to 2.4) | < 0.001 | 1.0 (0.6 to 2.3) | 0.7 (0.3 to 2.3) | 0.003 |
Values were considered significant at P < 0.05.
Represents GLS values for the specific layer evaluated by use of the vendor-specific and vendor-independent software, respectively.
Weighted mean was used, with the weighting greatest at the endocardium.
— Not applicable; values could not be determined by use of the software.
Reproducibility of GLS values obtained with each of the 3 software systems was good, as indicated by within-day and between-day intraobserver and interobserver CVs < 10% (Tables 2 and 3). However, reproducibility for systolic and diastolic SR was not always satisfactory, especially for late diastolic SR values, for which CVs reached 18.7%.
Intraobserver and interobserver variability of STE-derived longitudinal strain and SR for 6 dogs obtained with vendor 1-dependent software and vendor-independent software.
Intraobserver variability | ||||||
---|---|---|---|---|---|---|
Within-day CV (%) | Between-day CV (%) | Interobserver CV (%) | ||||
Variable | Vendor 1 | Independent | Vendor 1 | Independent | Vendor 1 | Independent |
GLS | ||||||
Transmural* and endocardial | 3.5 | 5.6 | 3.1 | 9.5 | 8 | 3.9 |
Systolic SR | 7.4 | 8.6 | 6.2 | 9.8 | 7.8 | 5.7 |
Early diastolic SR | 11.3 | 5.8 | 9.3 | 16.9 | 8.8 | 1 1.1 |
Late diastolic SR | 14.2 | 13.1 | 8.5 | 18.7 | 14.2 | 17.3 |
Weighted mean was used, with the weighting greatest at the endocardium.
Intraobserver and interobserver variability of STE-derived longitudinal strain and SR for 6 dogs obtained with vendor 2-dependent software and vendor-independent software.
Intraobserver variability | ||||||
---|---|---|---|---|---|---|
Within-day CV (%) | Between-day CV (%) | Interobserver CV (%) | ||||
Variable | Vendor 2 | Independent | Vendor 2 | Independent | Vendor 2 | Independent |
GLS | ||||||
Mid-myocardial | 2.8 | — | 5.5 | — | 6.8 | — |
Endocardial | 2.9 | 6.2 | 7.4 | 3.8 | 9.9 | 7.1 |
Systolic SR | 6.9 | 8.9 | 8.1 | 6.8 | 7.6 | 15.1 |
Early diastolic SR | 4.1 | 11.1 | 10.8 | 4.7 | 11.7 | 16.2 |
Late diastolic SR | 6.0 | 10.9 | 11.0 | 16.9 | 17.1 | 11.6 |
See Table 1 for remainder of key.
2-D STE analysis of images for the vendor 1 system
Comparison of most strain indices revealed significant differences among software systems, with absolute numbers for mean GLS values slightly higher for the vendor 1-specific software than for the vendor-independent software (Table 1). Evaluation of Bland-Altman plots revealed a mean difference of −0.9% (95% LOA, −6.5% to 4.7%; Figure 3). For systolic SR, the mean difference was −0.3 1/s (95% LOA, −1.5 to 0.9 1/s). For early diastolic SR, the mean difference was 0.1 1/s (95% LOA, −1.5 to 1.7 1/s). For late diastolic SR, the mean difference was 0.7 1/s (95% LOA, −0.5 to 1.9 1/s).
2-D STE analysis of images for the vendor 2 system
Comparison of each strain index revealed significant differences among software systems (Table 1). Absolute numbers for vendor 2-derived mid-myocardial GLS appeared to be slightly lower than results for vendor-independent-derived endocardial GLS. In contrast, mean vendor 2-derived endocardial GLS was higher than mean vendor-independent-derived endocardial GLS.
Evaluation of Bland-Altman plots revealed a mean difference between vendor 2-derived mid-myocardial GLS and vendor-independent-derived endocardial GLS of 1.3% (95% LOA, −4.4% to 7.0%; Figure 4). The mean difference between vendor 2-derived endocardial GLS and vendor-independent-derived endocardial GLS was −2.6% (95% LOA, −8.9% to 3.7%).
For systolic SR values, evaluation of Bland-Altman plots revealed a mean difference of −0.2 1/s (95% LOA, −1.4 to 1.8 1/s). For early diastolic SR, the mean difference was 0.7 1/s (95% LOA, −1.7 to 3.1 1/s). For late diastolic SR, the mean difference was 0.3 1/s (95% LOA, −0.7 to 1.3 1/s).
Discussion
To the authors' knowledge, the study reported here was the first in which investigators compared results of various 2-D STE software systems for longitudinal strain analysis in dogs with and without cardiac disease. Results indicated that despite use of good-quality images and a high frame rate, reproducibility was not uniform across software systems. When strain analysis was feasible, the reproducibility of GLS measurements was good and superior to that of SR for each software system. Finally, there was significant variability in GLS and SR measurements among the software systems.
In the present study, vendor-independent software could not reliably be used to analyze images attained from 11 of 55 (20%) dogs initially included in the vendor 1 cohort, despite inclusion of a prospective image acquisition protocol for 2-D STE strain analysis. A potential determining factor could have been the use of DICOM files for the vendor-independent system, which had a lower frame rate and image quality than did the raw image files used for the vendor-dependent software.7 A low frame rate can affect reliability of strain analysis by influencing temporal resolution, whereas poor image quality can impede accurate tracking of speckles by affecting spatial resolution.7 For speckle tracking analysis of echocardiographic images of humans, an acquisition frame rate > 30 Hz, and ideally approximately 50 Hz, is recommended.4 In the present study, frame rates > 60 Hz were selected because dogs typically have higher heart rates than humans, and the rate appeared to be appropriate, at a minimum, for raw format image analysis because all images could be analyzed with vendor-specific software. Considering that image quality appeared to be equally good in all the dogs included and that most of the images obtained with the ultrasound system of vendor 2 by use of the predetermined frame rate could be reliably analyzed with the vendor-independent software, another possible determinant could have been differences in technology of the ultrasound machines. In fact, image characteristics may differ among ultrasound machines on the basis of temporal and spatial resolution, filter settings, and other postprocessing effects.8 Heart rate can also affect accuracy of 2-D STE measurements, and high heart rates at conventional frame rates might result in underestimation of GLS.28 In a study29 conducted to assess the effects of heart rate on strain variables, values did not differ significantly when pacing rates between 120 and 180 beats/min were used and frame rate was increased with increasing heart rate. In the present study, frame rate was not adjusted on the basis of heart rate; instead, the frame rate was carefully maintained within the predetermined range; thus, it might be possible that this intervention resulted in inadequate frame rate for some of the images exported to the DICOM format. Reliable analysis can be performed on DICOM data compressed at 30 frames/s, but lower strain values may be observed.30 Although heart rate, frame rate, ultrasound technology, and data compression could all have affected strain assessment, the source of the dissimilar reproducibility of the vendor-independent analysis remains unclear.
When strain analysis was reproducible, intraobserver and interobserver agreement of GLS values was good for each software system, which is similar to results of studies of humans28 and dogs.12,13 Reproducibility of systolic SR also appeared to be acceptable because CVs were generally < 10%, except for interobserver variability in the vendor 2 cohort, for which the maximum value recorded was 15.1%. In contrast, diastolic SR indices had overall poorer reproducibility. Although each software system provided GLS values automatically, in most cases, they displayed SR measurements only after the observer placed a cursor over the global SR waveform at an appropriate time point, which potentially added variability. Different user interactions and semiautomatic user guidance may have a substantial impact on measurement variability among systems, particularly for SR data.6,8 Furthermore, time-dependent variables (eg, SR) are more difficult to assess than is deformation and require a higher frame rate.6 Regardless of the cause, GLS was the most reproducible longitudinal strain index in the present study.
Significant differences in strain variables were found among the 2-D STE software systems, and some results for individual dogs differed considerably. Unfortunately, the source of discordance could not be further defined because a criterion-referenced technique for deformation imaging (ie, sonomicrometry) was not available.7 Discrepancies among manufacturers are mainly attributed to differences in STE algorithms; there is a lack of published information on software validation, and most are not open-source algorithms.7 There have been overall improvements in variation among vendors since the creation of the EACVI-ASE Industry Strain Standardization Task Force,31,32 and it is hoped that these improvements will continue. However, considering that purchasing updated software might incur substantial costs, it is likely that, at least for veterinary settings, older versions of software will be in use for a considerable period during which variability will have to be taken into account. In the study reported here, repeated measurements of GLS obtained for the same dog with 2 software systems could differ by up to 5.6 to 6.3 strain units, which suggested a change of < 6.5% for the GLS could have been within measurement error for the various postprocessing software systems. This may have an important clinical impact because serial echocardiographic examinations to determine GLS could be used to track subclinical changes in LV function over time.31
Software systems can also be used to evaluate the various myocardial layers, which could represent an important source of variability. Each software system used in the present study automatically displayed a value defined as GLS, but this was sometimes derived from a specific layer rather than from the entire myocardium. For example, the vendor-independent software derived GLS from endocardial strain data (values from epicardial layers were excluded because the analysis was not considered reliable), which might have caused overestimation of strain values, compared with a software system that derived GLS from the myocardial midline.7 On the other hand, as stated by the manufacturer's instructions, vendor 1-specific software incorporated a weighted mean of the myocardial deformation across the myocardium in strain calculations, with the weighting greatest at the endocardium. Endocardial, midline, and epicardial strain values and mean values calculated over the entire cardiac wall can all potentially be obtained,28 and the presence of a longitudinal strain gradient from the endocardial layer to the epicardial layer has been established, which possibly is caused by increases in end-diastolic wall stress toward the endocardium and differences in coronary perfusion and metabolism between the layers.33 However, use of strain data obtained at different layers as the only explanation of variability among results for various software systems in the study reported here was not confirmed by comparing the endocardial measurements obtained with vendor 2-specific software and measurements obtained with the vendor-independent software, which also were derived from the endocardial layer. In fact, significant differences were found between vendor 2-derived endocardial GLS and vendor-independent-derived endocardial GLS, and Bland-Altman analysis provided a similarly wide LOA for all longitudinal strain values that were evaluated. Similarly, results of another study8 indicated moderate significant differences between endocardial GLS strain values among systems of 9 vendors. Furthermore, it is expected that endocardial strain values would be higher than transmural strain values,33 as detected for the vendor 2 cohort but in contrast to results detected for the vendor 1 cohort. On the other hand, comparison of vendor-dependent software by use of the entire myocardial wall and vendor-independent software by use of only endocardial strain (as was performed in the present study) resulted in small nonsignificant mean differences in peak strain between STE software systems in another study.7 In addition, comparison of 2 software programs, both of which analyzed full myocardial thickness, led to better agreement between global strain measurements than that found when comparing results for a system that determined mean transmural strain to another system that primarily analyzed endocardial strain.34 Therefore, results of between-layer strain comparability appeared to be conflicting. Because there is currently no evidence about the relative clinical value of strain measurements for various specific layers, endocardial, midline, epicardial, and transmural strain values can all be obtained, provided the layer analyzed is reported,28 and values derived from different ROIs are not used interchangeably without appropriate precautions.
The EACVI-ASE Industry Strain Standardization Task Force also advocates the use of end-systolic strain values as default variables to describe myocardial deformation.6 In the present study, peak strain measurements were chosen because these were provided automatically and therefore considered more likely to be used in clinical veterinary settings in the absence of specific guidelines.
The present study involved dogs with and without myocardial disease to ensure a broad spectrum of strain values. The LV function ranged from enhanced (eg, some dogs with myxomatous mitral valve disease at stage B2) to severely reduced (eg, some dogs with idiopathic dilated cardiomyopathy). However, LV function has not been found to significantly affect longitudinal strain variability,30 and the inclusion of both cardiovascularly healthy dogs and dogs with a variety of cardiac diseases was intended to provide useful results for clinical settings.
The present study had limitations. The number of dogs included was small and not uniform across groups, and there was a dissimilar distribution of dogs within disease category. Because it was a clinical study, it was not possible to control for the status of the animals included; at the same time, a broad range of strain values was desired. Therefore, a similar study conducted with a larger population of dogs or within a narrowly defined disease category could provide different results. A second limitation was that some of the affected dogs received medication, and some of the cardiac treatments that were administered could possibly have affected LV function and therefore strain variables. However, because a paired sample design was used and each dog served as its own control animal, with the same echocardiographic image analyzed for each dog, the inclusion of treated patients was allowed. A third limitation was that dogs underwent echocardiographic examination by use of a single vendor ultrasound system depending on their location, whereas they ideally should have undergone echocardiography by use of both ultrasound systems, which would have permitted us to make direct comparisons between the systems. Furthermore, the present study did not include regional strain analysis; this was intentional because one of the vendor-specific software systems used a segmentation model that differed from those of the other 2 systems.
Another limitation was the use of an early version of the vendor-dependent 2-D STE software, although there can be significant variability of strain measurements obtained by use of various versions of the same 2-D STE system. More specifically, higher GLS values have been obtained in healthy humans by use of one of the versions of the software system used in the present study, compared with values for a more recent version of that system,35 and improved agreement of GLS between the latest versions of the same software from different vendors has been reported.32
Finally, endocardial strain and transmural strain were directly compared, despite the differences known to exist between them that are related to the gradient between endomyocardium and epimyocardium. Ideally, only values obtained from the same ROI should be compared. However, commonly used software systems were used in the study reported here, and comprehensive layer-specific analysis was not always available; thus, the variables examined were primarily those provided by default. Therefore, results of the present study appeared to the authors to be useful because they could aid in assessing, as carefully as possible, strain indices obtained in clinical situations.
In the study reported here, reproducibility of analysis was not uniform across software systems, but when strain assessment could be performed, the reproducibility of GLS measurements was good and superior to that of SR. Because of significant variability in GLS and SR measurements obtained by use of the various software systems, it would appear necessary to be cautious when comparing values obtained for an individual dog to its previous measurements or to reference values that have been derived by use of other ultrasound machines, software, and even other versions of the same software.
Acknowledgments
Supported by an international mobility grant from the International School of Doctoral Studies of the University of Murcia and the Campus of International Excellence Mare Nostrum.
The authors thank Yolanda Martinez-Pereira for resident supervision at the University of Edinburgh, Alisdair Boag for language editing and proofreading, Darren Shaw for assistance with the statistical analysis, and Frane Ivasovic and Sara-Ann Dickson for technical assistance.
ABBREVIATIONS
ASE | American Society of Echocardiography |
CV | Coefficient of variation |
EACVI | European Association of Cardiovascular Imaging |
GLS | Global longitudinal strain |
LOA | Limits of agreement |
LV | Left ventricle |
ROI | Region of interest |
SR | Strain rate |
STE | Speckle tracking echocardiography |
Footnotes
iE33 ultrasound system, Philips Medical Systems, Andover, Mass.
Vivid E9, General Electric Medical Systems, Waukesha, Wis.
QLAB 2D strain software, version 9.0, Philips Medical Systems, Andover, Mass.
TomTec 2D cardiac performance analysis (2D CPA), Image-Arena, version 4.6, TomTec Imaging Systems, Unterschleissheim, Germany.
EchoPach PC, version 113, GE Vingmed Ultrasound AS, Horten, Norway.
IBM SPSS statistics for Windows, version 21.0, IBM Corp, Armonk, NY.
References
1. Chetboul V, Serres F, Gouni V, et al. Radial strain and strain rate by two-dimensional speckle tracking echocardiography and the tissue velocity based technique in the dog. J Vet Cardiol 2007;9:69–81.
2. Armstrong WF, Ryan T. Evaluation of systolic function of the left ventricle. In: Feigenbaum's echocardiography. 7th ed. Philadelphia: Lippincott Williams & Wilkins, 2010;123–157.
3. Armstrong WF, Ryan T. Specialized echocardiographic techniques and methods. In: Feigenbaum's echocardiography. 7th ed. Philadelphia: Lippincott Williams & Wilkins, 2010;39–66.
4. Blessberger H, Binder T. Non-invasive imaging: two-dimensional speckle tracking echocardiography: basic principles. Heart 2010;96:716–722.
5. Gorcsan J III, Tanaka H. Echocardiographic assessment of myocardial strain. J Am Coll Cardiol 2011;58:1401–1413.
6. Voigt JU, Pedrizzetti G, Lysyanski P, et al. Definitions for a common standard for 2D speckle tracking echocardiography: consensus document of the EACVI/ASE/Industry Task Force to standardize deformation imaging. Eur Heart J Cardiovasc Imaging 2015;16:1–11.
7. van Everdingen WM, Maass AH, Vernooy K, et al. Comparison of strain parameters in dyssynchronous heart failure between speckle tracking echocardiography vendor systems. Cardiovasc Ultrasound 2017;15:25.
8. Farsalinos KE, Daraban AM, Ünlü S, et al. Head-to-head comparison of global longitudinal strain measurements among nine different vendors. The EACVI/ASE inter-vendor comparison study. J Am Soc Echocardiogr 2015;28:1171–1181.
9. Amundsen BH, Helle-Valle T, Edvardsen T, et al. Noninvasive myocardial strain measurement by speckle tracking echocardiography: validation against sonomicrometry and tagged magnetic resonance imaging. J Am Coll Cardiol 2006;47:789–793.
10. Culwell NM, Bonagura JD, Schober KE. Comparison of echocardiographic indices of myocardial strain with invasive measurements of left ventricular systolic function in anesthetized healthy dogs. Am J Vet Res 2011;72:650–660.
11. Chetboul V, Serres F, Gouni V, et al. Noninvasive assessment of systolic left ventricular torsion by 2-dimensional speckle tracking imaging in the awake dog: repeatability, reproducibility, and comparison with tissue Doppler imaging variables. J Vet Intern Med 2008;22:342–350.
12. Dickson D, Shave R, Rishniw M, et al. Echocardiographic assessments of longitudinal left ventricular function in healthy English Springer Spaniels. J Vet Cardiol 2017;19:339–350.
13. Santarelli G, Talavera López J, Fernández del Palacio J. Effects of a combination of acepromazine maleate and butorphanol tartrate on conventional and two-dimensional speckle tracking echocardiography in healthy dogs. Am J Vet Res 2017;78:158–167.
14. Wess G, Keller LJM, Klausnitzer M, et al. Comparison of longitudinal myocardial tissue velocity, strain, and strain rate measured by two-dimensional speckle tracking and by color tissue Doppler imaging in healthy dogs. J Vet Cardiol 2011;13:31–43.
15. Pedro B, Stephenson H, Linney C, et al. Assessment of left ventricular function in healthy Great Danes and in Great Danes with dilated cardiomyopathy using speckle tracking echocardiography. J Vet Cardiol 2017;19:363–375.
16. Smith DN, Bonagura JD, Culwell NM, et al. Left ventricular function quantified by myocardial strain imaging in small-breed dogs with chronic mitral regurgitation. J Vet Cardiol 2012;14:231–242.
17. Suzuki R, Matsumoto H, Teshima T, et al. Clinical assessment of systolic myocardial deformations in dogs with chronic mitral valve insufficiency using two-dimensional speckle-tracking echocardiography. J Vet Cardiol 2013;15:41–49.
18. Zois NE, Tidholm A, Nägga KM, et al. Radial and longitudinal strain and strain rate assessed by speckle-tracking echocardiography in dogs with myxomatous mitral valve disease. J Vet Intern Med 2012;26:1309–1319.
19. Zois NE, Olsen NT, Moesgaard SG, et al. Left ventricular twist and circumferential strain in dogs with myxomatous mitral valve disease. J Vet Intern Med 2013;27:875–883.
20. Hamabe L, Kim S, Yoshiyuki R, et al. Echocardiographic evaluation of myocardial changes observed after closure of patent ductus arteriosus in dogs. J Vet Intern Med 2015;29:126–131.
21. Spalla I, Locatelli C, Zanaboni AM, et al. Echocardiographic assessment of cardiac function by conventional and speckle-tracking echocardiography in dogs with patent ductus arteriosus. J Vet Intern Med 2016;30:706–713.
22. Spalla I, Locatelli C, Zanaboni AM, et al. Speckle-tracking echocardiography in dogs with patent ductus arteriosus: effect of percutaneous closure on cardiac mechanics. J Vet Intern Med 2016;30:714–721.
23. Takano H, Fujii Y, Yugeta N, et al. Assessment of left ventricular regional function in affected and carrier dogs with Duchenne muscular dystrophy using speckle tracking echocardiography. BMC Cardiovasc Disord 2011;11:23.
24. Boon J. The two-dimensional echocardiographic exam. In: Veterinary echocardiography. 2nd ed. Oxford, England: Blackwell Publishing, 2011;37–99.
25. Boon J. The M-Mode and Doppler examination. In: Veterinary echocardiography. 2nd ed. Oxford, England: Blackwell Publishing, 2011;101–151.
26. Thomas WP, Gaber CE, Jacobs GJ, et al. Recommendations for standards in transthoracic two-dimensional echocardiography in the dog and cat. Echocardiography Committee of the Specialty of Cardiology, American College of Veterinary Internal Medicine. J Vet Intern Med 1993;7:247–252.
27. Atkins C, Bonagura J, Ettinger S, et al. Guidelines for the diagnosis and treatment of canine chronic valvular heart disease. J Vet Intern Med 2009;23:1142–1150.
28. D'hooge J, Barbosa D, Gao H, et al. Two-dimensional speckle tracking echocardiography: standardization efforts based on synthetic ultrasound data. Eur Heart J Cardiovasc Imaging 2016;17:693–701.
29. Suzuki R, Matsumoto H, Teshima T, et al. Influence of heart rate on myocardial function using two-dimensional speckle-tracking echocardiography in healthy dogs. J Vet Cardiol 2013;15:139–146.
30. Risum N, Ali S, Olsen NT, et al. Variability of global left ventricular deformation analysis using vendor dependent and independent two-dimensional speckle-tracking software in adults. J Am Soc Echocardiogr 2012;25:1195–1203.
31. Costa SP, Beaver TA, Rollor JL, et al. Quantification of the variability associated with repeat measurements of left ventricular two-dimensional global longitudinal strain in a real-world setting. J Am Soc Echocardiogr 2014;27:50–54.
32. Shiino K, Yamada A, Ischenko M, et al. Intervendor consistency and reproducibility of left ventricular 2D global and regional strain with two different high-end ultrasound systems. Eur Heart J Cardiovasc Imaging 2017;18:707–716.
33. Nagata Y, Chien-Chia Wu V, Otsuji Y, et al. Normal range of myocardial layer-specific strain using two-dimensional speckle tracking echocardiography. PLoS One 2017;12:e0180584.
34. Biaggi P, Carasso S, Garceau P, et al. Comparison of two different speckle tracking software systems: does the method matter? Echocardiography 2011;28:539–547.
35. Nagata Y, Takeuchi M, Mizukoshi K, et al. Intervendor variability of two-dimensional strain using vendor-specific and vendor-independent software. J Am Soc Echocardiogr 2015;28:630–641.