• Open Access

Left Ventricular Radial and Circumferential Wall Motion Analysis in Horses Using Strain, Strain Rate, and Displacement by 2D Speckle Tracking


  • Previously presented at the 26th Annual Forum of the ACVIM, San Antonio, TX, June 4–7, 2008.a

Corresponding author: Colin C. Schwarzwald, Dr med vet, PhD, Dipl ACVIM, Equine Department, Vetsuisse Faculty of the University of Zurich, Winterthurerstrasse 260, 8057 Zurich, Switzerland; e-mail: cschwarzwald@vetclinics.uzh.ch.


Background: Noninvasive assessment of left-ventricular (LV) function is clinically relevant, but is incompletely studied in horses.

Objectives: To document the feasibility, describe the techniques, and determine the reliability of 2D speckle tracking (2DST) for characterization of LV radial and circumferential wall motion in horses.

Animals: Three Standardbreds, 3 Thoroughbreds; age 8–14 years; body weight 517–606 kg.

Methods: Observational study. Repeated 2-dimensional echocardiographic examinations were performed in unsedated horses by 2 observers and subsequently analyzed by 2DST. Test reliability was determined for segmental and for averaged 2DST indices (including strain, strain rate, displacement, and rotation) by estimating measurement variability, within-day interobserver variability, between-day interobserver variability, and between-day intraobserver variability. Variability was expressed as coefficient of variation (percent) and the absolute value below which the difference between 2 measurements will lie with 95% probability.

Results: 2DST analyses were feasible in 16 of 18 echocardiographic studies. The automated tracking was accurate during systole but inaccurate during diastole. Reliability was higher for radial compared to circumferential measurements. For radial strain, radial systolic strain rate, and radial systolic displacement, the test-retest variabilities ranged between 2.4 and 33.1% for segmental and between 4.1 and 16.1% for averaged measurements.

Conclusions and Clinical Importance: Systolic radial motion of the LV at the chordal level could be reliably characterized in horses by 2DST. Circumferential measurements were less reliable. Diastolic measurements were invalid because of inaccurate tracking. The clinical value of LV wall motion analysis by 2DST in horses requires further investigation.


aortic valve closure (automatically determined)


aortic valve closure (manually determined)


British Standards Institution


coefficient of variation




radial peak systolic displacement


2D speckle tracking




circumferential peak strain




radial peak strain


radial peak systolic strain


instantaneous heart rate (for AVCa)


instantaneous heart rate (for AVCm)


left ventricle or left-ventricular


postsystolic index


region of interest


peak systolic rotation


strain rate


circumferential late-diastolic strain rate


circumferential early-diastolic strain rate


circumferential systolic strain rate


radial late-diastolic strain rate


radial early-diastolic strain rate


radial systolic strain rate


synchrony time index


within-subject standard deviation


time of aortic valve closure (automatically determined)


time of aortic valve closure (manually determined)


tissue Doppler imaging


time of peak radial strain

Ventricular function depends on a multitude of interrelated factors, including systolic and diastolic function, ventricular synchrony, and regional myocardial function.1–3 Echocardiography is the preferred method for clinical assessment of cardiac size and function in horses. Most of the echocardiographic indices used in clinical practice evaluate chamber dimensions and global left-ventricular (LV) function. Diastolic ventricular function, regional myocardial function, and ventricular synchrony are more difficult to quantify and therefore are rarely considered.4–8

Echocardiographic strain and strain rate imaging may provide additional indices for assessment of regional and global ventricular function beyond the conventional echocardiographic approach.9–11Strain (ɛ) is a measure of deformation of a myocardial segment, expressed as percentage of change from its original dimension.9,10 For normal myocardium, systolic strain is an analog of regional ejection fraction and reflects changes in stroke volume.9,12Strain rate (SR) is the temporal derivative of strain and describes the rate of deformation.9,10 Systolic SR reflects regional contractile function and seems to be relatively independent of heart rate and loading conditions in some studies.9,12

Strain and SR in people have been valuable in the diagnosis of subclinical myocardial disease, coronary artery disease, and myocardial involvement in noncardiac diseases, and in differentiation of hypertrophy caused by hypertension or cardiomyopathy.9,11 Strain and SR imaging can be further useful in valvular heart disease to detect subclinical systolic dysfunction and to understand the physiological impact of disease.9 Furthermore, it may allow objective assessment of myocardial performance during stress testing to identify stunned myocardium, acute ischemia, and myocardial infarction.9,13 Changes in magnitude of strain and SR caused by myocardial disease may be more striking than changes in time intervals.9 However, ventricular dyssynchrony may be of clinical and prognostic importance in people with heart failure, although the exact pathophysiological background and the clinical relevance of dyssynchrony are unknown in many affected patients.14,15

To our current knowledge, horses rarely suffer from clinically relevant myocardial ischemia, coronary artery disease, and myocardial infarction, and the prevalence of congestive heart failure is certainly lower than in people and small animals. However, myocardial diseases may be underappreciated in horses, at least in part because of the limitations of the currently used diagnostic methods.16 The clinical relevance of abnormal myocardial motion and ventricular dyssynchrony in horses is unknown to date. Therefore, there is a need to investigate novel echocardiographic methods such as strain and SR imaging, to better quantify myocardial wall motion abnormalities and ventricular dyssynchrony that could be suggestive of occult myocardial disorders.

Strain and SR can be calculated using a variety of echocardiographic techniques, including conventional M-mode imaging, tissue Doppler imaging (TDI), and 2D speckle tracking (2DST).2,9–11,17 In our pilot studies on the use of TDI in horses,18 the calculations of transmural velocity gradients, radial strain, and radial SR in the LV free wall were found to be highly unreliable (unpublished data). 2DST is a recent addition to advanced echocardiography systems that is based on conventional two-dimensional (2D) recordings. It analyzes the gray scale “speckles” as they move during contraction and relaxation and allows tracking of myocardial motion throughout the entire cardiac cycle. The 2DST method offers a Doppler-independent approach to the measurement of strain and SR that may overcome some of the limitations of M-mode-based and TDI-based strain measurements.2,9–11,17 Importantly, 2DST is independent of the angle of insonation and allows simultaneous measurement of segmental and global strain and SR in 2D.9,10 Furthermore, 2DST provides data on rotational motion, derived from circumferential strain at different levels of the left ventricle, that can be used to quantify ventricular systolic torsion (twist).19–21 Thus, if complex regional wall motion patterns are to be investigated, 2DST is considered a preferred method.10,22

The goal of this study was to demonstrate the feasibility, describe the techniques, and determine the reliability of transthoracic echocardiography for characterization of LV mechanical function in horses using 2DST. We hypothesized that 2DST methods can be applied to characterize radial and circumferential LV strain and SR as well as radial displacement, and rotation based on 2D-echocardiographic recordings of the LV obtained during routine echocardiograms in standing, adult horses. Specifically, we aimed at formulating preliminary recommendations for echocardiographic assessment of LV mechanical function in horses by 2DST that can be used in future experimental and clinical studies.

Material and Methods

Study Population

Six horses (4 geldings, 2 mares; 3 Standardbreds, 3 Thoroughbreds) aged 9.5 (8–14) years (median, range) and with a body weight of 543 (517–606) kg were studied. All horses were part of the teaching herd of The Ohio State University, College of Veterinary Medicine. They were considered healthy based upon physical examination, cardiac auscultation, 2-minute surface electrocardiogram (ECG), and 2D and M-mode echocardiographic studies. None of the horses was in athletic condition, and none of them received medications during the 2 weeks preceding entry into the study. The studies were approved by the Institutional Animal Care and Use Committee of The Ohio State University.


All studies were conducted in unsedated horses standing in a quiet room, restrained by an experienced handler. Transthoracic echocardiography was performed using a digital echocardiographb with a phased array transducerc at a frequency of 1.9/4.0 MHz (octave harmonics). The imaging depth ranged between 26 and 28 cm and the sector width was reduced 1 step from its maximum width, in order to achieve a frame rate of at least 50 frames/s in 2D imaging mode. A single lead ECG was recorded simultaneously. At least 3 representative, nonconsecutive cardiac cycles were recorded in each view and stored as cine-loops in digital raw data format for offline analyses.d Cycles immediately after a sinus pause or 2nd degree atrioventricular block were precluded.

Routine transthoracic 2D, M-mode, and color Doppler echocardiographic studies were performed to assess cardiac structures, valvular competence, chamber dimensions, and LV systolic function, using standard right parasternal long-axis and short-axis views.7,8,23 The examination also included M-mode recordings of aortic valve motion in short-axis view as well as 2D cine-loop recordings of the LV outflow tract in a long-axis view and of the LV in a short-axis view at the chordal level.

Offline data analyses were performed at the Vetsuisse Faculty of the University of Zurich, Switzerland. Measurements were performed in random order and with the operator blinded to signalment, echocardiographer, and previous measurements, using a dedicated software package.d All measurements were performed on the raw data recorded during the routine echocardiograms. Three cardiac cycles were analyzed and for each variable, the average of the 3 measurements was used for further analyses.

First, the time of aortic valve closure, which was used to delineate the end of ventricular systole during strain analyses, was measured manually (tAVCm). It was defined as the time interval between the peak of the electrocardiographic S wave and the closure point of the aortic valve identified on a short-axis M-mode recording of the aortic valve. In cases where the closure point of the aortic valve was not clearly identifiable on the M-mode recording, anatomical M-moded was applied to the 2D cine-loop of the LV outflow tract in the long-axis view, and cursor placement was adjusted to identify the point of aortic valve closure. In addition to tAVCm, the R-R interval of each analyzed cycle was measured and the corresponding instantaneous heart rate (HRAVCm) was calculated as 60,000/R-R interval.

LV wall motion by 2DST was derived from the short-axis images of the LV recorded at the level of the chordae tendineae of the mitral valve. In order to be included in the analyses, at least 1 complete cardiac cycle of adequate quality had to be available on each recording. The 2DST variables were measured as follows: First, the appropriate short-axis image was selected and the “Q-Analysis” moduled was started. A single heart cycle was then selected by moving the left and the right cursor, respectively, to the peaks of the S waves on the ECG. Each cycle was identified by the unique image-number assigned by the software and the 1st and the last frame of the cycle. The frame rate was recorded. Subsequently, the “2D Strain” moduled was started. The short-axis grayscale loops of the LV were analyzed using the “SAX-PM” option. The region of interest (ROI) was determined by tracing the endocardial border of the LV at end-systole. The tracing was started at midseptum and proceeded in a clockwise direction. The chordae were not included in the tracings. The ROI width was then adjusted, so that the entire myocardial thickness was covered throughout the cardiac cycle. Subsequently, the speckle tracking analysis was started. The software algorithm automatically divided the myocardium into 6 segments of equal size, performed the speckle tracking analysis, and provided confirmation of adequate tracking for each segment (Fig 1). The segments were preselected by the software based on regional wall motion analysis standards applied to human patients and were not adjusted for use in the horse. The quality of the tracking was visually assessed by the operator during motion playback. If necessary, the line tracing the endocardium was readjusted and the speckle tracking analysis was repeated until adequate segmental tracking was confirmed by the software. Six curve profiles were then obtained, corresponding to the average of each of the 6 myocardial segments (Fig 2A–F). The following variables were derived: Circumferential strain (ɛC), circumferential strain rate (SRC), radial strain (ɛR), radial strain rate (SRR), radial displacement (DR), rotation (Rot), and rotation rate. Aortic valve closure was automatically calculated and displayed by the software (AVCa).

Figure 1.

 Right-parasternal short-axis view at the level of the chordae tendineae. The 2DST software automatically divides the region of interest into 6 segments, termed “AntSept”, “Ant”, “Lat”, “Post”, “Inf”, and “Sept”. Note that the labels assigned to the segments are based on the human echocardiographic nomenclature and do not directly correspond to the equine anatomy. Hence, apart from slight individual variations, “AntSept” and “Ant” generally depict the interventricular septum, whereas “Lat”, “Post”, “Inf”, and “Sept” depict the cranial, lateral, and caudal LV free wall. The quality of tracking is automatically verified by the software and segments with appropriate tracking are marked with a green “V” at the bottom of the screen.

Figure 2.

 Trace screens of the 2DST software displaying the following information: Top left: 2D image with the segmented ROI and parametric color coding at the time of aortic valve closure. Bottom left: M-mode with parametric color coding. Right: Trace display for the selected parameter. The colors of the traces correspond to the colors of the segmented region of interest (as shown in Fig 1). The dotted line (where shown) indicates the instantaneous average of all segments at the respective time of the cardiac cycle. An ECG is plotted for timing purposes. The start and the end of the cycle (S waves) are marked on the ECG with yellow dots. The time of aortic valve closure (AVC*) is indicated by a green vertical line, dividing the cycle in its systolic and diastolic component. A—Circumferential strain. B—Radial strain. C—Circumferential strain rate. D—Radial strain rate. E—Radial displacement. F—Rotation. See Table 1 for remainder of key.

The measurements for strain (%), strain rate (1/s), displacement (mm), and rotation (deg) were performed on the “Results” screen of the 2D Strain software module,d using the default settings for spacial smoothing, temporal smoothing, and drift-compensation. Rotation rate was not measured due to the inability to identify distinct peaks on the curve profiles. For all other measurements, automated detection of the peak values was verified on the graphical display and manually corrected by the observer as necessary. The tracings were stored using the “Approve” command.

Strain measurements for each segment included circumferential and radial peak strain (peak strain independent of aortic valve closure time; termed ɛC and ɛR) and radial peak systolic strain (radial peak strain during systole, before or at the time of aortic valve closure; termed ɛR-sys), respectively. Where the highest value for radial strain occurred before or at the time of aortic valve closure, ɛR was identical with ɛR-sys. Where 2 peaks for radial strain were present and ɛR-sys was followed by a 2nd, higher peak after aortic valve closure, the 2nd peak was considered ɛR. Where only 1 peak occurring after aortic valve closure was present, ɛR-sys was defined as radial strain at the time of aortic valve closure and the 2nd peak was termed ɛR.

SR measurements for each segment included circumferential and radial peak systolic strain rate (SRC-sys and SRR-sys), peak early-diastolic strain rate (SRC-E and SRR-E), and peak late-diastolic strain rate (SRC-A and SRR-A). Further measurements for each segment included radial peak systolic displacement (DR-sys) and peak systolic rotation (Rotsys). The mean of the 6 segmental measurements of each variable was calculated to obtain indices of averaged strain, strain rate, displacement, and rotation. The cycle length (R-R interval) and the time interval from the S wave to AVCa (termed tAVCa) were measured. The corresponding instantaneous heart rate (HRAVCa) was calculated as 60,000/R-R interval. Lastly, the time of occurrence of peak radial strain (tɛR) of each segment was measured in each cycle and the synchrony time index (STIɛR), a measure of myocardial dyssynchrony, was calculated as the difference in tɛR from the earliest to the latest segment.24,25 Postsystolic motion was diagnosed where ɛR occurred after aortic valve closure.

Reliability of Echocardiographic Variables

All horses underwent repeated echocardiographic examinations by 2 experienced examiners, according to previously determined imaging guidelines. One echocardiographer (CCS) examined each horse twice at an interval of 2 days. On one occasion, a 2nd, independent echocardiographer (KES) examined each horse immediately before (3 horses) or after (3 horses) the other echocardiographer. All recordings were labeled with random codes, allowing subsequent offline measurements in a blinded fashion.

The intraobserver measurement variability was determined by a single observer (CCS) measuring the same 6 studies (1 study of each horse) repeatedly on 2 different days, thereby averaging the same set of 3 cardiac cycles for each variable. For determination of the interobserver measurement variability, a 2nd observer (AJB) measured the same cardiac cycles on the same 6 studies, independently of the 1st observer. For determination of the interobserver within-day variability, 1 observer (CCS) measured the 2 studies of each horse that were recorded consecutively on the same day by the 2 observers. The intraobserver between-day variability was determined by 1 blinded observer (CCS) measuring each horse's 2 studies that were recorded by CCS on different days. The interobserver between-day variability was determined by 1 blinded observer (CCS) measuring 2 studies of each horse that were recorded by the 2 observers 2 days apart; the studies were chosen so that 3 horses were examined by CSS first, and 3 horses were examined by KES first. All measurement were performed with the stored recordings in random order and with the observers blinded to signalment and previous measurements.

Data Analysis and Statistics

All statistical and graphical analyses were performed by standard computer software.e,f,g The test reliability was quantified using the within-subject variance for repeated measurements (residual mean square) determined by 1-way analysis of variance with the horses being the “groups”.26 The within-subject standard deviation (sw) was calculated as the square root of the residual mean square. Measurement variability and recording variability were reported in 2 ways: (1) The within subject coefficient of variation (CV) expressed as a percent value was calculated as CV =sw/mean × 100,26 in order to compare the reliability of the various variables within this study. The degree of variability was arbitrarily defined as follows: CV < 5%, very low variability; 5–15%, low variability; 16–25%, moderate variability; >25%, high variability.23 (2) In addition to the CV, the absolute value below which the difference between 2 measurements will lie with 95% probability was estimated following the British Standards Institution (BSI) recommendations: BSI = 1.96 ×√2 ×sw= 2.77 ×sw.26 The BSI was reported to provide a clinically applicable measure of variability, hence an absolute value that allows comparison with measured changes in echocardiographic variables on a case-by-case basis.

Summary statistics (mean ± SD) of each variable were calculated using the 1st study of each horse (n = 6) and were reported for comparison. The prevalence of postsystolic motion in the study population (as defined above) was also assessed based on the 1st study of each horse. Agreement between the tAVCa and the tAVCm was determined based on the same studies by a Wilcoxon signed-rank test and Bland-Altman statistics; the corresponding heart rates (HRAVCa and HRAVCm) were compared by a Wilcoxon signed-rank test. The level of significance was defined as α= 0.05.


The results of the routine 2D and M-mode echocardiography of the population studied have been previously reported.18 Of the 18 echocardiograms recorded from the 6 horses, 16 recordings were of adequate quality to allow 2DST analyses; 2 recordings were excluded from analysis due to insufficient image quality or incomplete recordings of the cardiac cycle. The frame rate of the recordings ranged from 59 to 73 frames/s. Automated confirmation of adequate tracking was achieved in all segments of all analyzed cycles by both observers. Subjective assessment of tracking by the observer revealed that tracking was accurate during systole but appeared inaccurate (too slow) during early diastole in all recordings, despite confirmation of adequate tracking by the software. Examples of the tracings given by the analysis software are shown in Figure 2A–F. The tracings for strain, SR, displacement, and rotation, respectively, showed a consistent pattern among all horses.

The tAVCa was 481 (396–518) ms (median, range) and the tAVCm was 499 (378–530) ms; they did not differ significantly (P= 0.44). Agreement of the tAVCa and the tAVCm was characterized by a mean bias of −10 ms and 95% limits of agreement between −68 and +48 ms. The heart rates during the respective recordings did not differ significantly (P= 0.56). Using tAVCa to define end-systole, postsystolic radial thickening was present in 5 out of the 6 horses on their 1st echocardiogram, affecting 80 (74%) of 108 segments in 10 (56%) of 18 analyzed cardiac cycles.

The reliability data of all segmental and averaged variables are summarized in Tables 1 and 2. To facilitate interpretation of the data, the reliability data of strain, SR, displacement, and rotation are also graphically displayed in Figure 3. Figure 4 shows the pooled variability data, grouped by myocardial segment. Figure 5 provides a graphical overview on the pooled variability data of segmental and averaged variables, grouped by variable.

Table 1.   Reliability of segmental 2DST variables of circumferential and radial LV wall motion of 6 healthy horses
VariablesSegmentaMean ± SDbMeasurement variabilityWithin-day variabilityBetween-day variability
n = 6n = 6n = 5n = 5n = 5
  • a

    See Figure 1.

  • b

    b Summary statistics based on 1st study of each horse.

  • c

    c Variables considered being sufficiently reliable for further use in experimental and clinical studies.

  • CV (%) = percent coefficient of variation; BSI = absolute value below which the difference between 2 measurements will lie with 95% probability (following the British Standards Institution).

  • ɛC, circumferential peak strain; ɛR, radial peak strain; ɛR-sys, radial peak systolic strain; SRC-sys, circumferential systolic strain rate; SRC-E, circumferential early-diastolic strain rate; SRC-A, circumferential late-diastolic strain rate; SRR-sys, radial systolic strain rate; SRR-E, radial early-diastolic strain rate; SRR-A, radial late-diastolic strain rate; DR-sys, radial peak systolic displacement; Rotsys, peak systolic rotation.

ɛC (%)AntSept−24.19 ± 8.414.83.2528.317.4223.114.7024.414.8617.711.88
Ant−23.03 ± 7.407.54.7320.012.1518.811.2827.618.5017.610.79
Lat−15.21 ±
Post−5.81 ± 6.4264.913.36203.719.16237.818.87218.629.91211.113.77
Inf−12.01 ± 4.1425.09.1370.516.5523.48.1250.418.9743.715.66
Sept−19.78 ± 5.846.53.5866.126.1114.98.2330.316.5825.115.05
ɛR (%)cAntSept59.96 ± 14.854.47.346.09.799.415.6715.424.5011.720.53
Ant48.53 ± 9.974.25.649.713.267.610.0222.230.418.811.65
Lat47.70 ± 15.247.910.3910.513.8324.530.4422.231.9423.931.13
Post56.34 ± 18.6812.418.9313.620.7033.148.8820.433.8826.542.43
Inf63.18 ± 17.4812.020.5013.422.8526.645.1422.439.5517.732.66
Sept64.52 ± 15.608.314.579.416.2313.924.2921.136.2311.822.29
ɛR-sys (%)cAntSept57.14 ± 13.404.16.536.19.5612.820.0515.222.5815.425.02
Ant44.11 ±
Lat44.51 ±
Post53.05 ± 18.2712.217.6013.218.5635.247.1524.336.3627.337.51
Inf59.69 ± 15.5211.618.7911.117.7626.341.6126.442.4814.524.02
Sept61.09 ± 13.628.013.416.911.3912.019.9523.737.8910.017.55
SRC-sys (1/s)AntSept−1.30 ±
Ant−1.03 ± 0.3311.00.3334.60.9617.00.4521.80.6617.10.45
Lat−0.84 ±
Post−0.64 ± 0.18114.81.69109.22.50135.01.5069.11.21106.91.19
Inf−0.87 ± 0.4338.81.0538.10.8356.21.2045.31.2438.90.91
Sept−1.14 ±
SRC-E (1/s)AntSept2.02 ± 0.4714.50.8630.31.4925.61.2926.01.2421.61.09
Ant1.97 ± 0.497.30.3927.71.3525.21.1922.21.1929.91.34
Lat1.93 ± 0.5010.40.5422.21.0934.61.5529.21.4118.30.70
Post2.09 ± 0.9212.80.7722.31.1744.82.2440.41.7534.41.49
Inf1.50 ± 0.5312.70.5523.40.8527.81.0823.00.8325.30.94
Sept1.61 ± 0.529.60.4431.21.1911.10.4826.51.1117.70.82
SRC-A (1/s)AntSept1.06 ±
Ant0.84 ± 0.327.00.1646.90.9442.30.7015.00.3540.50.69
Lat0.64 ± 0.388.10.1479.51.1826.80.3940.70.7330.30.47
Post0.65 ±
Inf0.61 ± 0.1315.80.2799.91.2927.20.4056.90.9862.90.90
Sept1.00 ± 0.339.90.2974.41.5030.80.7145.71.2340.50.88
SRR-sys (1/s)cAntSept1.37 ±
Ant1.50 ±
Lat1.67 ±
Post1.77 ±
Inf1.76 ±
Sept1.58 ±
SRR-E (1/s)AntSept−1.94 ± 0.635.50.2913.20.6818.20.9418.10.8514.00.72
Ant−1.62 ± 0.868.50.3819.30.8624.91.1130.01.1121.60.96
Lat−1.05 ± 0.5217.60.5432.91.0838.71.1936.91.0826.20.85
Post−0.99 ± 0.3528.10.8337.21.2129.50.8738.91.2325.30.80
Inf−1.29 ± 0.3524.00.8630.41.1725.40.9429.51.1122.90.88
Sept−1.60 ± 0.339.80.4319.60.8820.10.8919.20.8018.70.84
SRR-A (1/s)AntSept−1.27 ±
Ant−1.25 ± 0.417.10.2432.00.9827.20.855.90.2023.00.72
Lat−1.44 ± 0.636.80.2740.61.4326.00.9320.20.8225.80.96
Post−1.56 ± 0.759.40.4142.11.6231.41.1931.51.4628.71.18
Inf−1.61 ± 0.6911.30.5041.91.6835.71.4033.01.5630.11.30
Sept−1.50 ± 0.4711.90.5067.42.1233.51.2423.91.0329.51.17
DR-sys (mm)cAntSept14.79 ±
Ant15.29 ± 2.592.
Lat19.66 ± 2.694.22.334.22.2515.77.6817.99.4720.09.53
Post21.16 ± 3.706.73.999.85.4420.310.9315.08.3618.19.59
Inf18.77 ± 4.646.93.6614.46.9120.79.9312.25.9020.910.31
Sept16.31 ± 4.354.52.0510.74.5817.87.6112.75.4917.27.81
Rotsys (deg)AntSept3.69 ±
Ant6.64 ± 1.442.50.4512.
Lat7.64 ± 1.577.61.5519.53.8310.72.2826.65.4916.93.74
Post4.68 ± 1.819.51.2450.35.9973.07.6045.65.7768.07.54
Inf2.04 ± 2.9026.71.6093.45.04309.210.39207.39.59294.58.67
Sept2.29 ± 3.2123.81.5380.34.54162.18.92153.18.35147.96.50
Table 2.   Reliability of averaged 2DST variables of circumferential and radial LV wall motion of 6 healthy horses
VariablesMean ± SDaMeasurement variabilityWithin-day variabilityBetween-day variability
n = 6n = 6n = 5n = 5n = 5
  • a

    Summary statistics based on 1st study of each horse.

  • b

    b Variables considered being sufficiently reliable for further use in experimental and clinical studies.

  • CV (%) = Percent coefficient of variation; BSI = Absolute value below which the difference between two measurements will lie with 95% probability (following the British Standards Institution).

  • ɛC, circumferential peak strain; ɛR, radial peak strain; ɛR-sys, radial peak systolic strain; SRC-sys, circumferential systolic strain rate; SRC-E, circumferential early-diastolic strain rate; SRC-A, circumferential late-diastolic strain rate; SRR-sys, radial systolic strain rate; SRR-E, radial early-diastolic strain rate; SRR-A, radial late-diastolic strain rate; DR-sys, radial peak systolic displacement; Rotsys, peak systolic rotation; tAVCa, automatically calculated time of aortic valve closure; STIɛR, difference in time to peak radial strain from the earliest to the latest segment (synchrony time index). See Table 1 for remainder of the key.

ɛC (%)−14.25 ± 2.499.23.7328.39.3811.95.1430.211.8816.56.30
ɛR (%)b48.61 ± 10.716.38.428.711.6015.323.3716.121.9512.016.69
ɛR-sys (%)b45.66 ± 9.696.47.976.88.4113.519.0120.826.008.110.09
SRC-sys (1/s)−0.83 ± 0.0617.00.4030.60.7020.80.4920.90.4822.40.45
SRC-E (1/s)1.59 ±
SRC-A (1/s)0.69 ±
SRR-sys (1/s)b1.38 ±
SRR-E (1/s)−1.21 ± 0.3212.50.4221.40.7517.90.7126.30.8410.20.35
SRR-A (1/s)−1.23 ± 0.438.70.3035.41.0528.10.9937.01.2826.40.83
DR-sys (mm)b15.14 ± 1.684.11.757.73.0912.65.6712.04.8515.76.08
Rotsys (deg)3.85 ±
tAVCa (ms)476 ± 450.680.911.737.81105.8787.8109
STIɛR (ms)104 ± 5435.29941.514650.114840.113452.4155
Figure 3.

 Graphical illustration of the pooled reliability data of segmental (A) and averaged (B) 2DST indices of LV function, grouped by type of variability. The different types of variability are displayed on the horizontal axes; the coefficients of variation (CV) are displayed on the vertical axes (log scale). Variability is classified as very low (CV < 5%), low (CV = 5–15%), moderate (CV = 16–25%), and high (CV > 25%).

Figure 4.

 Graphical illustration of the pooled reliability data of segmental 2DST indices of LV function, grouped by segment (see Fig 1). The segments are displayed on the horizontal axes; the corresponding coefficients of variation (CV) are displayed on the vertical axes (log scale). (A) Shows all variables, (B) shows the variables of radial systolic LV function (ɛR, ɛR-sys, SRR-sys, DR-sys) only.

Figure 5.

 Graphical illustration of the pooled reliability data of segmental (A) and averaged (B) 2DST indices of LV function grouped by index. The 2DST indices are displayed on the horizontal axes; the corresponding coefficients of variation (CV) are displayed on the vertical axes (log scale).


The results of this investigation reveal that 2DST of radial and circumferential motion of the LV is feasible using standard 2D echocardiographic recordings of standing, unsedated, adult horses. Selected 2DST indices have sufficient test-retest reliability to justify further investigations to the use of 2DST-based wall motion analyses for objective quantification of systolic LV function in horses. Specifically, the indices of systolic radial LV function are sufficiently reliable to potentially aid in sequential testing of LV systolic function over time, eg, to monitor progression of disease or response to treatment.

The importance of reliability data as well as details on calculation and interpretation of indices of reliability have been discussed elsewhere.23 The results of this study show that, with only a few exceptions, most 2DST variables have very low to moderate intraobserver measurement variability, indicating that image quality and measurement guidelines were overall sufficient (Fig 3). However, the interobserver measurement variability was markedly higher and ranged from low to high, with only 1 variable in the “very low” range. Similar to the results of a previous study on TDI,18 this finding emphasizes the importance of extensive observer training and strict application of the measurement guidelines. Generally, it is advised that 1 single operator should perform sequential measurements to minimize measurement error.

The within-day, interobserver variability ranged between low and high. The overall higher variability compared with the intraobserver measurement variability was primarily attributed to the error introduced by a 2nd operator performing the examination, while some degree of short-term biological variability may also have played a role. This finding again emphasizes the importance of well-defined imaging guidelines and adequate operator training.

The between-day intraobserver variability and between-day interobserver variability reflected the combined effects of biological variability, recording variability, and measurement error. These 2 measures of test reliability may be most important for clinical applications.23 In general, true alterations in echocardiographic variables caused by disease, drugs, or interventions in an individual patient would have to be larger than the changes simply caused by day-to-day variability (as quantified by the BSI). Thus, the true clinical value of the method under investigation not only depends on the variability of measurements but also depends on the magnitude of echocardiographic changes seen in a clinical setting.

Comparison of the variability of pooled data among the 6 myocardial segments (Fig 4) showed that the overall variability of measurements was somewhat lower in the septal and cranial wall segments (termed “Ant Sept”, “Ant”, “Lat”) compared with the segments of the lateral and caudal LV wall (termed “Post”, “Inf”, “Sept”). As indicated previously, the wall segments preselected by the software were based on human regional wall motion analysis standards. While it is possible to adjust the borders and override the system, this was not done in this study. The biological relevance of this segmental analysis has not been confirmed in horses and should be considered as a potential limitation of the methodology applied to this species.

Further evaluation of the variability using pooled data revealed that segmental and averaged indices of radial systolic LV function (ɛR, ɛR-sys, SRR-sys, and DR-sys) performed overall better than variables of circumferential systolic LV function (ɛC, SRC-sys, and Rotsys) and diastolic variables (SRR-E, SRC-E, SRR-A, and SRC-A; Fig 5). Also, the validity of diastolic SR variables has to be questioned due to the fact that the tracking quality was judged by the operators to be insufficient during the diastolic phases of the cardiac cycle. As a consequence, the 2DST method may only be suitable for the assessment of systolic radial LV dysfunction. Conversely, the findings of this study do not justify the future use of 2DST indices of circumferential LV function and of diastolic LV function in clinical or experimental settings.

Evaluation of ventricular dyssynchrony by 2DST can be achieved by visual assessment of the trace display and the color-coded M-mode display (Fig 2) and by calculation of a variety of synchrony indices.24,25,27–29 Derivation of these indices often requires some subjective assessment of which peaks to measure, a task that is not always objective and straightforward.11,14 Currently, there is no single synchrony index that is considered optimal for assessment of ventricular dyssynchrony in people, and studies suggest that combining dyssynchrony data from different methods may be of additive value.28 In this study we chose to investigate the STIɛR, an index that can easily be calculated based on segmental time intervals.24,25,29 The results showed that the reliability of the STIɛR was poor. This finding is in agreement with recent human studies that suggested insufficient reliability of analysis of dyssynchrony for clinical use.11,30 The high variability of segmental time intervals may, in part, be related to technical issues. More specifically, smoothing filters and the relatively low frame rate at which 2DST loops are recorded may lead to a loss of detail in the strain and SR curves, which may reduce the validity and reliability of temporal variables.9,10 Other factors possibly leading to poor reliability of the STIɛR include biological variability and inadequate measurement guidelines and recording techniques. In this study, we used the timing of the segmental ɛR to calculate the STIɛR, aiming at assessing the degree of dyssynchrony independent of the end of systole. The current study design included strict, objective measurement guidelines that did not allow any subjective judgment of strain curves to decide which peak strain values to include in the calculation of the STI. It is possible that calculating the STI based on ɛR-sys instead of ɛR would improve its reliability. However, by doing so, postsystolic myocardial motion occurring after aortic valve closure (which seems to be common in healthy horses) would be ignored, and myocardial dyssynchrony may therefore be missed in some patients. Nonetheless, it seems inevitable to interpret synchrony indices in association with the subjective assessment of the graphical display of segmental myocardial motion. Whether the application of different, partly subjective measuring guidelines, the use of higher frame rates, or the assessment of longitudinal rather than radial motion would improve the reliability of myocardial dyssynchrony assessment in horses is currently unknown. In depth evaluation of other synchrony indices would have been beyond the scope of this study. Further studies are needed to investigate the clinical relevance and the best diagnostic approach to ventricular dyssynchrony in horses.

Aortic valve closure indicates the end of ventricular mechanical systole. Estimation of the time of aortic valve closure during the cardiac cycle is important to distinguish systolic peak strain from maximum peak strain. These are different in the presence of postsystolic motion, the degree of which can be quantified as postsystolic index (PSI; calculated as postsystolic increment divided by systolic strain).9 While postsystolic motion is reported in over 30% of myocardial segments in healthy people, a high PSI (ie, >35%) in conjunction with reduced systolic strain occurs with ischemia, LV hypertrophy, or as a passive phenomenon in dyskinetic segments.9,11,13 The concept of postsystolic motion has not been investigated in horses. Using our admittedly arbitrary definition of postsystolic motion, its prevalence in this supposedly healthy study population was high. However, in this study we did not attempt to grade the degree of postsystolic motion or to investigate its causes or its clinical relevance. Therefore, it remains uncertain whether the definition of postsystolic motion used in this study would be clinically useful. Nonetheless, the tAVCa, which is automatically calculated by the analysis software, was shown to be reliable and was in good agreement with tAVCm, which is measured manually. Automated timing of aortic valve closure on each cardiac cycle may in fact be superior to an averaged estimation of tAVCm, as the latter cannot account for variations of heart rate between different cycles.31 Therefore, we suggest that tAVCa should be used if the clinical relevance of postsystolic motion in horses were to be investigated in future studies.

This study has some limitations. The data obtained from our study will not be unconditionally applicable to all clinical situations due to differences in patient population, equipment, machine settings, analysis software, observer experience, and image quality. We limited our investigations to radial and circumferential wall motion and to human-defined segments of the LV that can be easily assessed using the standard LV short-axis views in horses. We did not study longitudinal strain, SR, and displacement. In people, longitudinal strain is most commonly used, while measurement of radial strain from tissue velocity data is thought to be unsuitable for clinical use. One of the reasons for the preference of longitudinal strain over radial strain in humans is that the optimal intersite distance (also referred to as strain length) required for TDI-based strain measurements (12 mm) cannot be achieved in a ventricle with normal wall thickness, leading to marked signal noise.9 However, this restriction does not hold for adult horses, in which the normal thickness of the LV free wall ranges between 20 and 35 mm.7,32 Furthermore, SR imaging by 2DST—as opposed to TDI—is based on a different concept that does not depend on local strain length. Therefore, radial strain and SR by 2DST may be well suitable for use in this species and certainly warrants further investigation.

The 2DST method comes along with its own intrinsic limitations, independent of the study design. First, attention to technical detail is of crucial importance during image acquisition and 2DST analyses. The 2D echocardiographic loops need to be of sufficient quality and free of artifacts. Second, speckle tracking has limitations in sampling rate. The recommended frame rate for recording of images ranges between 40 and 90 frames/s.d10 Higher frame rates are associated with a high level of noise and reduced lateral resolution, whereas lower frame rates risk loss of correlation and poor tracking because of excessive displacement of the speckles.9,10 However, at the recommended frame rates, shorter events like the isovolumic phases may disappear and peak values may be reduced due to undersampling. This is particularly relevant for measuring peak values in early diastole and during the brief isovolumic phases. Systolic strain and systolic SR are less frame rate sensitive due to the relatively lower wall motion velocities during systole compared with diastole.18,10,33 In the present study, the limitation in sampling rate, possibly leading to undersampling of rapid wall motion events, was likely responsible for the poor tracking observed during the diastolic phase of the cardiac cycle. A 3rd potential limitation of 2DST is that the longitudinal motion of the heart base leads to considerable through-plane motion when imaging the LV in short axis, which may impair tracking accuracy for radial and circumferential events by current 2D algorithms. Nonetheless, acknowledging all the limitations related to study design and methodology, 2DST, used in conjunction with conventional 2D, M-mode, and Doppler echocardiography, may add to the understanding of LV mechanical function and may aid in the diagnosis of global and segmental LV systolic dysfunction in horses at rest and during physical or pharmacological stress testing.

In conclusion, we were able to show that LV mechanical function can be evaluated noninvasively using 2DST-based wall motion analysis applied to standard 2D echocardiographic recordings of the LV in a right-parasternal short-axis view at the level of the chordae tendineae. We suggest that 2DST-based indices may be useful for assessment of systolic LV performance and LV synchrony in horses. However, our imaging guidelines and recommendations should be regarded as preliminary and will need validation in clinical practice. Further studies will be required to investigate the use of other echocardiographic short-axis and long-axis planes for 2DST analysis and to establish reference values for specific patient populations. Finally, the clinical value of 2DST to assess disease-related alterations in LV systolic function at rest and after stress-testing, and their relation to severity of disease, exercise capacity, and prognosis need to be investigated.


aSchwarzwald CC, Schober KE, Bonagura JD: Echocardiographic characterization of left ventricular radial and circumferential wall motion in horses using strain, strain rate, and displacement by 2D speckle tracking: Methodology and reliability. J Vet Int Med 2008;22:759 (Abstract)

bGE Vivid 7, BTO4, GE Medical Systems, Milwaukee, WI

cM3S phased array transducer, GE Medical Systems

dEchoPAC Software v6.1.2, GE Vingmed Ultrasound A/S, Horten, Norway

eMicrosoft Office Excel 2003, Microsoft Corporation, Redmond, WA

fSigmaStat v3.01, SPSS Inc, Chicago, IL

gGraphPad Prism v5.00 for Windows, GraphPad Software, San Diego, CA