Strain Estimation of the Murine Right Ventricle Using High-Frequency Speckle Tracking Ultrasound

Abstract

Right ventricular (RV) strain measurements from ultrasound via speckle tracking techniques are being used more frequently as a non-invasive diagnostic tool for a variety of cardiopulmonary pathologies. However, despite the clinical utility of ultrasound RV strain measurements, quantification of RV strain in rodents remains difficult due to unique image artifacts and non-standardized methodologies. We demonstrate here a simple approach for measuring RV strain in both mice and rats using high frequency ultrasound and automated speckle tracking. Our results show estimated peak RV free wall longitudinal strain values in mice (n=15) and rats (n=5) to be −10.38±0.4% and −4.85±0.42%, respectively (mean±SEM). We further demonstrated estimation of 2D Green-Lagrange strain within the RV free wall, with longitudinal components estimated at −5.7±0.48% in mice and −2.1±0.28% in rats. These methods and data may provide a foundation for future work aimed at evaluating murine RV strain levels in different disease models.

Introduction:

Rapid and accurate evaluation of the right ventricle (RV) is crucial to better understanding the pathological progression of many diseases that affect cardiopulmonary function. However, measurement of right heart function in preclinical small animal models using ultrasound is limited by sternum and lung anatomical image artifacts, non-standardized echocardiographic imaging windows, and differences in analysis methods (Bernardo, et al. 2017, Espe, et al. 2020, Kimura, et al. 2015, Kohut, et al. 2016). Utilizing small animal models provides an essential platform to validate methodologies and develop treatments in a controlled, preclinical setting. There is thus a need to improve methods that allow for the measurement of RV strain in mice and rats.

To date, ultrasound quantification of RV strain has been utilized in the clinic to assess outcomes of patients with conditions such as pulmonary embolism, heart failure, arrhythmogenic right ventricular cardiomyopathy, acute respiratory distress syndrome, aortic coarctation, pulmonary hypertension, congenital heart defects, and other pathologies (Labombarda, et al. 2020, Lemarié, et al. 2020, Malik, et al. 2020, Padiyath, et al. 2013, Sciaccaluga, et al. 2020, Shukla, et al. 2018, Trivedi, et al. 2020). A recent study by Li and colleagues also suggests that RV longitudinal strain is a powerful predictor of mortality in patients with COVID-19 with 94.4% sensitivity and 64.7% specificity (Li, et al. 2020). Given the global prevalence of these diseases and pathologies, it is essential to continue improving methods of diagnosis and treatment.

Murine models of heart disease have been and continue to be of paramount importance to the development of our understanding and treatment of cardiopulmonary disease. In particular, genetic, chemical, and surgical murine models of RV pathology have already been reported in literature (Ciuclan, et al. 2011, Maarman, et al. 2013, Stenmark, et al. 2009). While ultrasound analysis of the right ventricle has been limited with these models in the past, continued improvements in technology and methods may allow us to utilize these models to their full potential by improving image quality and reproducibility. Additionally, a standardized approach will help with reproducibility and yield reliable metrics for longitudinal monitoring of disease progression; particularly important in studies of pulmonary hypertension or the cardiopulmonary sequela of COVID-19.

While a small number of groups have successfully estimated RV strain in rats using speckle tracking ultrasound (Bernardo, et al. 2017, Kimura, et al. 2015), the presented work is to our knowledge the first demonstration of measuring RV strain in mice. This technical note outlines one approach for obtaining and analyzing RV strain data in mice and rats using commercially available high frequency ultrasound imaging and speckle tracking software. Additional analysis techniques and data comparison are also described using automated feature tracking data outputs. This non-invasive method also allows for a reduction in total animals used in longitudinal studies of RV function as each animal can be imaged at multiple time points without necessitating sacrifice for data collection. This type of non-invasive analysis of RV function may be valuable in rodent models of pulmonary hypertension, heart failure, right-sided myocardial infarction, and perhaps even in elucidating the cardiac complications of COVID-19 (Espe, et al. 2020, Kimura, et al. 2015, Zhu, et al. 2019). Additional benefits include cost-savings compared to other non-invasive imaging modalities such as magnetic resonance or computed tomography. While we present here one process to obtain and analyze RV strain data from wild-type mice and rats, we hope this study lays a foundation for more refined techniques to be developed and a wider array of preclinical RV disease models to be investigated.

Materials/Methods:

Animals

For this study, twenty-five-week-old female wild-type CD-1 mice (n=5; Charles River, Wilmington, MA), ten-week-old female C57BL/6J (B6) mice (n=5; Jackson Laboratories, Bar Harbor, ME), eleven-week-old male B6 mice (n=5; Jackson Laboratories, Bar Harbor, ME), and six-week-old male Sprague Dawley wild-type rats (n=5; Charles River, Wilmington, MA) were used for imaging. The Purdue University and FUJIFILM Visual Sonics Inc. Institutional Animal Care and Use Committees approved all animal procedures. Prior to ultrasound imaging, we induced anesthesia using 3-4% isoflurane in 1.5 L/min O₂. During ultrasound imaging, we maintained heart rate between 450-550 BPM in mice and 350-450 BPM in rats by modulating anesthesia level with 1.5-2.5% isoflurane in 1.5 L/min O₂. Temperature was maintained between 35-37°C using a heat-modulated imaging stage (Vevo Imaging Station, FUJIFILM VisualSonics Inc., Toronto, Ontario, Canada) and rectal temperature probe, with electrodes for ECG monitoring.

Ultrasound Imaging

To prepare the mice and rats for imaging, we placed each animal in a supine position and applied depilatory cream to remove hair from the ventral thorax. Mice and rats were imaged using a Vevo 3100 high-resolution small animal ultrasound system (FUJIFILM VisualSonics Inc.). To image the rats, we used a 21 MHz center frequency linear array ultrasound transducer (15-30 MHz bandwidth; MX250). For mice, we used a 40 MHz center frequency linear array ultrasound transducer (25-55 MHz bandwidth; MX550D). The probes for both mice and rats were positioned immediately lateral to the sternum at approximately a 30° angle from midline in order to obtain a modified parasternal long axis view of the right ventricle. The stage was titled an additional 30° laterally to optimize the substernal view of the RV (Figure 1). Probe depth and angle were adjusted to limit rib and lung shadows that can obstruct the view of the RV free wall. B-mode data was acquired across 300 frames for speckle-tracking analysis. Imaging frame rates were 239-403 fps with the 40 MHz center frequency probe and 102-153 fps with the 21 MHz center frequency probe.

Figure 1.

Imaging Setup. (a,d) Right ventricular imaging of wild-type C57BL/6J (B6) mouse (a) and wild-type Sprague-Dawley rat (d) in a supine position. Stage and probe tilted for optimal substernal imaging. (b, e) Representative RV long-axis B-mode ultrasound image for mouse (b) and rat (e), showing RV free-wall epicardial border (red dotted line) and RV free-wall endocardial border (blue dotted line) used in analysis and rib shadow artifact (white arrows). Scale bar = 1 mm. RV (right ventricle), LV (left ventricle), RA (right atrium), LA (left atrium), PA (pulmonary artery). (c) Depiction of the modified RV long axis B-Mode image acquisition technique used to obtain RV free wall motion for analysis.

Speckle Tracking Analysis

We performed speckle tracking analysis using the Vevo Strain software (Vevo LAB v3.2.6, FUJIFILM VisualSonics Inc.). Three consecutive cardiac cycles from the modified parasternal long-axis B-Mode videos were chosen for analysis using the anatomical M-mode setting to limit respiratory variations in RV wall motion (Figure 2). This setting allows for visualization of wall motion abnormalities at a particular cross section of the 2-dimmensional image across time, thus enabling exclusion of cardiac cycles affected by respiratory artifacts. Utilizing the Free Curve software tool, a trained researcher manually placed 6-9 points along the endocardial free wall border at one time point within the three isolated cardiac cycles. Points were placed along the RV free wall such that they were relatively equidistant from each other, while also avoiding regions where rib or lung shadowing artifacts were observed (Figure 2). An initial execution of speckle tracking resulted in estimations of velocity, displacement, strain, and strain rate at each point. Afterwards, wall motion tracking was manually adjusted frame by frame by a trained reviewer as necessary. Following the endocardial RV free wall analysis, corresponding points were placed on the epicardial RV free wall and the aforementioned analysis was repeated accordingly. Using the tracked points from both myocardial wall boundaries, we resampled strain curves to a normalized cardiac cycle-interval so that data across animals could be directly compared, and then produced representative curves for each animal by averaging across all tracked points.

Figure 2.

Speckle-tracking analysis. (a) CD-1 mouse modified long axis B-Mode image. Dashed arrow indicates section for anatomical M-Mode analysis. Red dashed line indicates RV epicardial free wall and blue dashed line indicates RV endocardial free wall. (b) Anatomical M-mode with ECG tracing (top) and selection of three consecutive cardiac cycles for analysis^† which avoids respiratory artifacts during inspiration*. (c) Endocardial and (e) epicardial point placement, yellow dots indicate manually placed points, green path lines indicate point movement throughout cardiac cycles selected for analysis, white boxed images enhanced for path line visualization. (d) Endocardial and (f) epicardial strain software output blue strain curves superimposed on anatomical M-Mode and ECG tracing (green). Scale bars = 1 mm.

Segmental and Full-wall Strain Analysis

As a supplement to the strain estimations produced by Vevo Strain, speckle tracking-based displacement data was used to manually calculate RV strain both along each border and across the myocardium. In this approach, we assumed uniform deformations along each border to mitigate any tracking irregularities that might cause differences in strain along the RV. To enforce this assumption, a new set of points was derived at each time-point using a custom MATLAB script (MathWorks, Natick, MA, USA) by fitting a spline through the Vevo Strain points and then repositioning these points such that there were equal arc lengths between them. Displacement information was then rederived using these new points to calculate the linear, or “engineering”, strain from both Lagrangian (ε_Lag; eq.1) and Eulerian (ε_Eul; eq.2) frames of reference:

$${\epsilon }_{Lag}=\frac{\sum \Delta L}{L_0}$$ (1)

$${\epsilon }_{Eul}=\sum \frac{\Delta L}{L}$$ (2)

where L is the length between segments, ΔL is the change in length from one time point to the next, and L₀ is the length of a segment at t = 0 seconds.

Using the aforementioned equidistant points, each set of four points (i.e., two consecutive epicardial points and the corresponding two endocardial points) was used to define a quadrilateral segment from which full-wall strain tensors could be estimated. As mentioned previously, the points created for each analysis are resampled from the area of interest such that there are six equidistant points along the border of interest and these points are then matched with corresponding points on the opposite border. For each time point and quadrilateral segment, the 2-dimmensional version of the deformation gradient tensor (F) was derived (Humphrey, et al. 1990, Ligas, et al. 2019) using an affine transformation matrix between the point locations of the quadrilateral at t=0 and the current time-point. The deformation gradient (F) is calculated from the shift in the quadrilaterals and therefore is not dependent on initial shape of the quadrilateral. The deformation gradient tensor was then used to estimate both the Green-Lagrange strain tensor (E_2D; eq.3) and the Almansi-Eulerian tensor (e_2D; eq. 4) using the following equations:

$${\mathit{E}}_{2\mathit{D}}=\frac{1}{2}({\mathit{F}}^{\mathit{T}}\mathit{F}-\mathit{I}),{\mathit{E}}_{2\mathit{D}}=\left[\begin{array}{cc}\hfill {\mathrm{E}}_{\text{zz}}\hfill & \hfill {\mathrm{E}}_{zr}\hfill \\ \hfill {\mathrm{E}}_{\text{rz}}\hfill & \hfill {\mathrm{E}}_{\text{rr}}\hfill \end{array}\right]$$ (3)

$${\mathit{e}}_{2\mathit{D}}=\frac{1}{2}(\mathit{I}-(\mathit{F}{\mathit{F}}^{\mathit{T}}{)}^{-1}),{\mathit{e}}_{2\mathit{D}}=\left[\begin{array}{cc}\hfill {\mathrm{e}}_{\text{zz}}\hfill & \hfill {\mathrm{e}}_{\text{zr}}\hfill \\ \hfill {\mathrm{e}}_{\text{rz}}\hfill & \hfill {\mathrm{e}}_{\text{rr}}\hfill \end{array}\right]$$ (4)

where I is the identity matrix, E_zz and e_zz are longitudinal strain components, E_rr and e_rr are radial strain components, and E_zr, E_rz, e_zr, and e_rz are shear strain components. The longitudinal components of the Green-Lagrange (E_zz) and Almansi-Eulerian (e_zz) strains for each mouse and rat were then compared to the linear and speckle tracking-based longitudinal strain estimates.

Statistical Analysis

Prior to statistical comparisons, data from each group was checked for normality (Anderson Darling Test; α<0.01). Groups were considered statistically different when compared using a one-way analysis of variance (ANOVA) with p<0.05. As there was no statistically significant group differences in longitudinal strain (p=0.85), strain rate (p=0.60), peak velocity (p=0.79), or maximum displacement (p=0.61), results from each mouse group (i.e., female CD-1, female B6, and male B6 mice) were pooled into a cohort for subsequent analyses (Table 1). All data are shown as mean ± standard error of mean (SEM). As there were slight variations in heart rates between animals, graphical strain estimation across a cardiac cycle was accomplished by first normalizing the cardiac cycle for each animal. A 150 point linearly interpolated curve was then overlaid on each curve and the mean and 95% probability intervals of a t-distribution were calculated at each point by comparing the curve in each group of animals being analyzed.

Table 1:

Speckle tracking displacement, velocity, and strain metrics. Estimated speckle tracking parameter comparison between female CD-1 mice (n=5), female C57BL/6J (B6) mice (n=5), male B6 mice (n=5) and pooled results for combined mouse group (n=15; mean±SEM).

RV Free Wall:	Combined Mice (n=15)	Female CD-1 Mice (n=5)	Female B6 Mice (n=5)	Male B6 Mice (n=5)	p-value ANOVA
Long. Strain - Endo (%)	−10.38±0.40	−10.66±0.09	−10.07±0.98	−10.40±0.81	0.85
Long. Strain - Epi (%)	−8.67±0.48	−8.65±0.58	−9.48±0.91	−7.86±0.95	0.42
Strain Rate - Endo (1/s)	3.29±0.15	3.29±0.21	3.10±0.29	3.49±0.30	0.60
Strain Rate - Epi (1/s)	2.82±0.19	2.46±0.25	2.69±0.22	3.32±0.40	0.15
Peak velocity - Endo (cm/s)	1.06±0.05	1.11±0.11	1.04±0.08	1.03±0.09	0.79
Peak velocity - Epi (cm/s)	0.76±0.04	0.76±0.07	0.77±0.04	0.74±0.11	0.94
Peak Displacement - Endo (cm)	0.29±0.02	0.31±0.02	0.28±0.02	0.27±0.04	0.61
Peak Displacement - Epi (cm)	0.19±0.02	0.23±0.03	0.18±0.01	0.17±0.04	0.23

Results/Discussion:

Mouse RV Strain Estimates Consistent Between Ages, Sexes, and Genetic Strains

Table 1 includes measured values for RV free wall peak longitudinal strain, strain rate, wall velocity, and displacement for mice using the standardized speckle tracking method. Additionally, we examined a regional strain metric that determined the peak average strain for various segments of the RV free wall. Finally, we calculated the full-wall strain tensor in both the Almansi-Eulerian frame of reference as well as the Green-Lagrangian frame of reference and compared the peak averaged regional longitudinal strain component to our previous measures (Figure 3). The speckle tracking average peak strain estimate values for both the RV endocardial free wall and the RV epicardial free wall were −10.38±0.40% and −8.67±0.48% respectively. Our segmental and full-wall estimates of average Eulerian peak linear strain for the RV endocardial free wall and RV epicardial free wall were −6.47±0.62% and −4.86±0.64% respectively. Finally, the longitudinal component of the full-wall Almansi-Eulerian strain tensor was −6.73±0.78% and the longitudinal component of the Green-Lagrange strain tensor was −5.68±0.48%.

Figure 3.

Mouse Strain comparison. (a) CD-1 mouse RV long-axis B-Mode image, red dotted line indicates RV free wall epicardial border, blue dotted line indicates RV free wall endocardial border used in analysis, white scale bar = 1mm. (b) speckle tracking longitudinal strain curve estimates for a normalized cardiac cycle (mean±95%CI; n=15). (c) Average peak longitudinal strain values from speckle tracking (ST) estimation. (d) Graphical representation of segmental strain analysis. (e) Average segmental Eulerian strain curves for endocardial (orange) and epicardial (magenta) RV free wall across a normalized cardiac cycle. (f) Average peak longitudinal strain values from Eulerian segment strain estimation. (g) Graphical representation of strain tensor calculation for right ventricular free wall (h) average longitudinal component strain curves of Almansi-Eulerian (e₁₁, dark blue) and Green-Lagrange (E₁₁, green) strain tensors over a normalized cardiac cycle. (i) Average peak longitudinal component strain values from Almansi-Eulerian and Green-Lagrange strain tensors. Data are shown as mean±SEM.

From these results we found that we can reliably estimate physiologically relevant RV free wall strain values in mice. Additionally, we found that our RV estimated parameters were not significantly different between mouse groups with different age (10-25 weeks), sex, or genetic strain (CD-1 vs. B6), or when repeated measures were taken using the same imaging data set (Supplemental Figure 1). While the absolute values of RV wall motion parameters presented in this work may need additional validation, we have the ability to consistently estimate in vivo strain non-invasively.

Estimated Strain Values in Rats

In our rat experiments, we similarly determined average peak baseline values for longitudinal strain, strain rate, wall velocity, and displacement for the RV free wall by using a standardized method of speckle tracking ultrasound (Table 2). We also examined segmental and full wall strain metrics as described previously (Figure 4). The speckle tracking average peak strain estimate values for both the RV endocardial free wall and the RV epicardial free wall were −4.85±0.42% and −3.88±0.25% respectively. The average peak segmental strain estimates from the Eulerian frame of reference for the RV endocardial free wall and RV epicardial free wall were −2.55±0.54% and −1.79±0.59% respectively. Finally, the longitudinal component for the average peak full-wall strain estimate of the Almansi-Eulerian tensor was −2.41±0.27% and the longitudinal component of the Green-Lagrange tensor was −2.092±0.28%.

Table 2:

Estimated speckle tracking displacement, velocity, and strain metrics for Sprague Dawley (SD) rats (n=5; mean±SEM).

RV Free Wall:	Male SD Rats (n=5)
Long. Strain - Endo (%)	−4.85±0.42
Long. Strain - Epi (%)	−3.88±0.25
Strain Rate - Endo (1/s)	1.28±0.13
Strain Rate - Epi (1/s)	1.25±0.21
Peak velocity - Endo (cm/s)	0.82±0.15
Peak velocity - Epi (cm/s)	0.51±0.09
Peak Displacement - Endo (cm)	0.30±0.05
Peak Displacement - Epi (cm)	0.17±0.04

Figure 4.

Rat strain comparison. (a) Sprague Dawley rat RV long-axis B-Mode image, red dotted line indicates RV free wall epicardial border, blue dotted line indicates RV free wall endocardial border used in analysis, white scale bar = 1mm. (b) Speckle tracking longitudinal strain curve estimates for a normalized cardiac cycle (mean±95% CI; n=5). (c) Average peak longitudinal strain values from speckle tracking (ST) estimation. (d) Graphical representation of segmental strain analysis. (e) Average segmental Eulerian strain curves for endocardial (orange) and epicardial (magenta) RV free wall across a normalized cardiac cycle. (f) Average peak longitudinal strain values from Eulerian segment strain estimation. (g) Graphical representation of strain tensor calculation for right ventricular free wall (h) average longitudinal component strain curves of Almansi-Eulerian (e₁₁, dark blue) and Green-Lagrange (E₁₁, green) strain tensors over a normalized cardiac cycle. (i) Average peak longitudinal component strain values from Almansi-Eulerian and Green-Lagrange strain tensors. Data are shown as mean±SEM.

While literature of RV strain in rats is limited, there are several studies that have explored speckle tracking echocardiography (STE) applications. One study by Kimura et al. evaluated strain in an RV heart failure rat model, reporting RV longitudinal strain measurements in control animals of −8.3±2.8% (mean ± SD) (Kimura, et al. 2015). In this study, conventional parasternal long axis and apical 4-chamber views were used, though the authors recognized that obtaining these views was challenging due to the small echocardiographic window and noted issues with limited reproducibility. That study also used an ultrasound transducer with a lower center frequency of 11.5 MHz (Vivid 7, GE Healthcare, Waukesha, WI, USA) compared to the transducer with a 21 MHz center frequency used in our study (FUJIFILM VisualSonics Inc, Toronto, CA). After the images were obtained, the analysis was performed using an automated speckle-tracking software (EchoPac, GE Healthcare) with a similar workflow to that used in our study.

In the Study by Bernardo, et al. a modified parasternal and a short axis right ventricular outflow tract (RVOT) ultrasound views of the right ventricle were employed to overcome artifact limitations in the echocardiographic window (Bernardo, et al. 2017). Similar to our study, Bernardo, et al. used these views in rats to maximize visualization of the RV free wall. In our study, we found that a similar modification to the parasternal long axis view enabled a large portion of the RV free wall to be visualized and analyzed (Figure 1c.). The view we use in our study is similar to the RVOT view used in the study by Bernardo, et al. though our imaging window is shifted slightly such that a larger portion of the RV free wall in the long axis is visible.

This study also employed a lower frequency (6-14MHz) ultrasound transducer (Vivid E9, GE Vingmed Ultrasound AS, Horten, Norway) for image acquisition. While the imaging windows and acquisition methods were slightly different from our work, this study used a similar approach to analyze strain with manual tracing using a speckle tracking software (Echopac, version 2011, GE Healthcare). This study showed that RV strain analysis was feasible, however, they did not report a quantified strain metric due to the variability in their imaging data.

Espe et al. took a slightly different approach and examined RV strain in rats using cardiac magnetic resonance (CMR). In particular, phase contrast imaging was used to examine RV strain metrics in left ventricular infarct models of rats. Post-processing analysis was conducted using pixel-wise mapping of myocardial motion in MATLAB (MathWorks, Natick, MA, USA) rather than an automated speckle tracking algorithm. In this study, peak longitudinal strain in control rats was found to be −15% [interquartile range: −13.8% to −18.8%] (Espe, et al. 2020). These reported values and ranges are consistent with our strain estimates in mice; however, they are substantially higher than our measured values in rats. Differences in image acquisition, animal size, age, and analysis methods may account for discrepancies in quantified strain metrics. For example, significant differences have been shown in left-ventricular global longitudinal strain measurements when comparing between imaging system vendors (Farsalinos, et al. 2015). Additional limitations for comparing these studies include variation in image quality between animals, pathologic and experimental differences in the models used, standardization of echocardiographic views, and differences in imaging frequencies and modalities.

A portion of this difference might also be explained by a slight underestimation of strain using our particular method. Applying our strain estimation method to the left ventricular free wall, we observed average peak longitudinal strain in the left ventricle of a wild-type black 6 mouse group to be −11.14±0.81% (mean±SEM) for the endocardial free wall and −9.53±0.48% for the epicardial free wall (Supplemental Figure 2). Although our method is used to evaluate regional strain of the free wall and not a typical global longitudinal strain (GLS) metric, these values are still smaller than typical basal and mid- left ventricular free-wall regional peak strain values (−15% to −20%) in a healthy mouse (Bauer, et al. 2011, De Lucia, et al. 2019).

While it may be inappropriate to directly compare strain measurements across studies with different image acquisition and data analysis methods, measurements obtained using the proposed procedures should be reliable when studying and comparing right ventricular function changes longitudinally. In addition, this non-invasive method for RV strain estimation may still provide a useful tool for studying small animal models where other functional metrics are lacking. Further work will be needed for more comprehensive analysis and comparison.

Importance of RV Strain Estimation in Rodent Models

As this report focused on a method for obtaining baseline metrics of RV strain, we focused our analysis on three small groups of wild-type mice and one group of wild-type rats for comparison. Speckle tracking analysis was completed by one user to ensure consistency between groups. We found that our RV estimated parameters were not significantly different between our mouse groups with different age (10-25 weeks), biological sex, or genetic strain (CD-1 vs. B6). Though we did not anticipate large differences between these groups, the consistency is remarkable considering that image acquisition was performed by three different trained individuals at two different facilities. While the effects of biological sex, genetic strain, age, or imaging personnel might demonstrate to be more significant with a larger sample size, our initial results suggest that the proposed methods yield acceptable precision despite these differences.

The focus of this study was to determine the feasibility of measuring strain and other biomechanical parameters in the RV non-invasively using a relatively simple and quick method. While additional metrics such as RV ejection fraction and cardiac output can be used to assess RV function, these measurements are heavily reliant on geometric assumptions and are not always reported in 2D echocardiographic analysis of rodents (Kohut, et al. 2016). The globular shape of a mouse heart in addition to its substernal location makes it difficult to visualize the full RV. The use of 3D echocardiography or CMR may make the estimation of RVEF and cardiac output more feasible (Wang, et al. 2019), however the 2D focus presented in this study, while being relatively quick and easy, is not ideal for assessing these global metrics.

As this work focused on developing methods for RV strain estimation in healthy, wild-type rodents, future work will be needed for the application of this method to small animal models of cardiac disease. For example, this type of non-invasive analysis of RV function may be valuable in rodent models of pulmonary hypertension, heart failure, right-sided myocardial infarction, and perhaps even in elucidating the cardiac complications of COVID-19 (Espe, et al. 2020, Kimura, et al. 2015, Zhu, et al. 2019). We hope that this technical report can provide a foundation for future studies where functional metrics of RV function in small animals are limited.

Non-invasive Imaging and Speckle Tracking Strain Analysis Considerations

While an apical four chamber view would traditionally be used to assess RV function clinically (Phoon and Turnbull 2016), obtaining reliable RV wall motion data from four chamber views in rodents is notably more difficult due to its anterior location, complex shape, and sometimes prominent lung and rib shadowing artifacts (Kohut, et al. 2016, Rottman, et al. 2007). Although care was taken to limit image artifact from the ribs as described previously, significant shadowing was persistently observed. These artifacts were compensated for by placing boundary points only on portions of the RV free wall that were unaffected.

Using the speckle tracking software’s free curve tool allowed us to estimate only longitudinal strain (i.e., strain measured tangentially to the curve drawn by the user), thus radial or circumferential strains were not measured. The speckle tracking software used in this analysis has the capability of full-wall strain analysis, though it is optimized for the left ventricle where it expects defined continuous outlines of opposing walls. As it is very difficult to obtain a continuous view of both the RV free wall and RV septal wall in rats and mice, this particular full-wall analysis was not used given that it would require modification of proprietary tools for use in the RV. Additionally, while radial strain measurements are theoretically possible by combining endocardial and epicardial speckle tracking, and using our MATLAB-derived analysis for full-wall strain, we did not report radial strain as it is not considered a reliable metric in the thin-walled RV (Badano, et al. 2018).

Segmental and Full-wall Strain Analysis

The method used for strain analysis yielded slightly different estimates. For example, segmental estimation of strain from the Eulerian perspective in mice was significantly smaller (66.5%, endo; 61.0%, epi) than that of the speckle tracking estimate (p<0.001, endo; p<0.001, epi). This difference may be due to the uniform distribution and/or deformation assumptions made along each border in the segmental approach, compared to speckle tracking that allows each point to be tracked independently. For example, while care is taken to within the software to distribute points manually in an equidistant manner, this can sometimes be imprecise due to the need to avoid image artifact as the software is prone to error when portions of the wall are not tracked through an artifact. Another reason, as mentioned previously, is the deformation assumptions we impose to calculate our deformation tensor F and full-wall strain tensor. We perform these standardization techniques (equal arc length, deformation assumptions) in a post-processing manner in order to impose consistency in our calculations, however, we recognize that it is not a perfect metric and relies on a uniform distribution assumption. Because of these differences, it is important when estimating RV strain in small animals that proper control cohorts be included in the study design. We also observed slight differences when comparing the average peak longitudinal component of the Green-Lagrangian and Almansi-Eulerian 2D strain tensors, though these differences were not statistically significant in either mice (p=0.26) or rats (p=0.49). While these additional strain metrics have limitations, one strength of the presented method is the ability to measure both linear strain for the endocardial and epicardial surfaces as well as derive a full-wall 2-dimensional strain tensor. As with any method, consideration of these limitations and benefits in the context of study design will be important for optimal results.

Conclusion:

This study highlights a method for reliable RV strain analysis of mice and rats. These data suggest that right ventricular estimated parameters were not significantly different between mice with different age (10-25 weeks), biological sex, or genetic strain (CD-1 vs. B6). We also found that RV strain estimated in rats provides a basis for comparison with existing studies. This method of strain estimation allows for longitudinal assessments in the same animal for the characterization of strain differences over time or between control and diseased models. Our study provides initial RV strain and wall motion mechanical observations for a homogeneous population of wild-type mice and rats, laying a foundation for future studies of right ventricular pathologies using rodent models of cardiopulmonary disease.

Supplementary Material

1

Figure S1. Right Ventricular average peak longitudinal strain comparison and repeatability for endocardial free wall (Endo) and epicardial free wall (Epi) from speckle tracking echocardiography. A) Repeated measures for three different groups of wild-type mice; Female B6 (n=5), Male B6 (n=5), Female CD-1 (n=5). B) Three repeated independent strain measurements on Female B6 imaging data set (n=5). C) Repeated data collection and analysis for mouse groups at two different ages; Female B6 (n=5; 10 weeks vs 22 weeks) and Male B6 (n=5; 12 weeks vs 22 weeks). Bars are shown as mean±SEM. There are no significant differences between time points for either group.

2

Figure S2. Free-curve strain method estimation in the endocardial (endo) and epicardial free wall of the right ventricle (RV) and left ventricle (LV). Age-matched (8-10 weeks, n=5 RV, n=5 LV) female C57BL/6J mice were imaged for the analysis. Scatter points show biological replicates. Bar shown as mean±SEM.