To the best of the authors’ knowledge, this study is the first to compare transthoracic shear wave elastography (SWE) recordings throughout the entire cardiac cycle to invasive pressure-volume (PV) measurements for evaluating the myocardial stiffness dynamics in vivo. A moderate feasibility of the SWE-method was shown (success rate of 54%) and wave speed estimations varied 19.4% in diastole and 20.0% in systole on average. By performing a robust estimate of wave speed, this study linked diastolic wave speed to a pressure-volume loop-derived index of operational myocardial stiffness, i.e. dP/dV. Operational myocardial stiffness takes into account the stiffening effect due to increased ventricular pressure, stressing the importance of hemodynamic loading for diastolic wave speed. The ratio of systolic and diastolic wave speed correlated to pressure-volume loop-derived indices of contractility, i.e. end-systolic pressure (ESP) and preload-recruitable stroke work (PRSW). Measuring this ratio might thus provide a non-invasive index of contractility during I/R injury.
Feasibility of transthoracic SWE
The feasibility of SWE depends on three important factors: the amplitude of the acoustically induced vibrations, the quality of the images in which the vibrations are tracked, and the quality of the tracking algorithm itself to cope with underlying tissue motion. Transthoracic application of SWE to the heart poses extra difficulties as (i) cardiac phased array probes have a relatively small footprint and thus limited excitation strength and image quality, and (ii) cardiac tissue is moving relatively fast, challenging tracking algorithms but also push beam settings for powerful shear wave excitation throughout the cardiac cycle. This explains in part our moderate success rate of 54% (7/13 pigs). Some of the SWE acquisitions in the other 6 ‘unsuccessful’ pigs did reveal shear wave propagation but were not included in our analysis as there were only consistent SWE results for maximal 2 out of the 6 interventions. Especially at the start of our pig experiment, there were more ‘unsuccessful’ pigs as the experimental set-up gradually improved (use of a new generation probe and repositioning of the pigs at the beginning to obtain the best possible B-mode image quality in terms of contrast and SNR). Future work should investigate the feasibility of SWE in a systematic manner while varying the set-up and sequence settings, but this was outside the scope of this work.
Our reported feasibility is higher than the success rate of our dynamic SWE experiments in open-chest pigs 17, where a success rate of 32% was reported. It is also slightly higher than the success rate of 41% reported by Kakkad et al. 18, who used another technology (acoustic radiation force based imaging) to investigate qualitatively the dynamic stiffness variations across the cardiac cycle. The success rate increases tremendously when applying SWE at one time point during the cardiac cycle (end-diastole) according to previous clinical studies: 77.5% in healthy children 19 and 91.3% in healthy adults 20. Recently, some preliminary results have been reported on the dynamic stiffness variations in healthy children, but feasibility rate of these measurements is currently unknown 21. Technical improvements are currently ongoing to increase the feasibility of SWE: optimization of filters to suppress background motion 18, improvements in the imaging sequence, such as ultrafast harmonic coherent compound imaging 22, or development of a new ultrasound transducer with a better mechanical focus for shear wave excitation 23.
Next to the challenges to induce and observe the shear wave in the myocardium, there are also challenges to provide a robust estimate of the wave speed due to its variability (averaged variability of ± 20% in systole and diastole in this study). The variability in wave speed estimation might arise from the dependence of the wave propagation on the ventricular morphology, the stress state (ventricular pressures) and the myocardial fiber orientation 24. Furthermore, different acquisitions or heartbeats contribute to the natural physiological variability of the wave speed. Also, experimental settings such as the selected protocol, the number of observers and the selected wave speed estimator will affect wave speed estimation. We accounted for part of this variability by performing wave speed estimation for the total 10 M-mode lines across the septal wall drawn by 2 observers and subsequently fitting a piecewise linear model to the speed data of multiple heartbeats and acquisitions (see Fig. 2). Wave speed estimation was performed manually, which was time-consuming and subjective, and should be automated in future work. Furthermore, we believe that combining the linear regression fit with automation of wave speed estimation and M-mode line positioning are essential for an accurate and robust wave speed estimation, but also for clinical implementation.
Myocardial operational stiffness
Diastolic wave speed has been suggested as measure for myocardial stiffness by several studies through estimating the significance level of an infarcted vs. non-infarcted group in comparison to the significance of changes in other stiffness parameters: the stiffness constant of the end-diastolic stress-strain relation 6,7 and the end-diastolic pressure-volume relation (EDPVR) 5. Our study confirms this earlier research: the change in the stiffness constant β of the EDPVR increased after the ischemia period (0.073 vs. 0.045 1/ml; p = 0.06) and increased even further after the reperfusion period (0.087 vs. 0.045 1/ml; p < 0.05), which was reflected in the change of diastolic wave speed after ischemia injury (2.0 vs. 1.3 m/s; p < 0.05) and reperfusion injury (2.5 m/s). Furthermore, our work showed a significant correlation between diastolic wave speed and stiffness constant β in Fig. 6c (R = 0.50; p = 0.04). Literature does not report any correlations between diastolic wave speed and stiffness parameters, however a clinical study in hypertrophic cardiomyopathy patients 20 showed a significant correlation between diastolic wave speed and fibrosis markers in cardiac magnetic resonance: T1 post-contrast (R = 0.595; p = 0.01) and late gadolinium enhancement (R = 0.80; p < 0.01).
The exponential coefficient β of the EDPVR is typically considered as index for diastolic chamber stiffness 14, however it does not provide insights into the stiffness dynamics. Operational chamber stiffness, where a change in pressure is considered relative to a change in volume (or dP/dV), allows to assess the local slope of the EDPVR at a specific filling pressure and thus increases when pressure increases due to the non-linear ESPVR. Therefore, we considered dP/dV next to stiffness constant β in this study in order to describe any stiffness changes due to loading and/or intrinsic stiffness alterations. As tabulated in Table 1, all interventions – hemodynamic or I/R injury – significantly altered the operational stiffness dP/dV. Diastolic wave speed significantly altered only after ischemia injury (see Fig. 5). It did show a significant positive correlation with dP/dV for both types of interventions (R = 0.57; p < 0.01 for loading and R = 0.68; p < 0.01 for stiffness interventions). Furthermore, the correlation was even stronger when considering all interventions (R = 0.75; p < 0.001 in Fig. 8a). This highlights that diastolic wave speed depends on ventricular loading and intrinsic stiffness and thus rather measures operational myocardial stiffness. The difference in the slopes of the fitted linear regression curves in Fig. 6b (0.80 vs. 0.35) suggests that diastolic speed is more sensitive to changes in intrinsic characteristics than changes in loading. This would be beneficial for clinical evaluation of intrinsic myocardial properties, or evaluation of loading if changes in stiffness can be ruled out based on the patient profile.
The results of our statistical analysis on the group level suggests a preload independence of diastolic wave speed, which is in correspondence with other studies 2,7, but our linear regression results demonstrate that loading does have an effect on diastolic wave speed, but to a lesser extent than the intrinsic properties. These observations can be partly explained by differences in the considered pressure change amongst studies (largest pressure drop in this study, i.e. 26.1 mmHg, but in the same range as observed in diseased human hearts 25), and the low sample size of our study. Furthermore, the effect of material nonlinearity can also be observed from the changes of the wave speed as a function of time during diastole in Fig. 2 and Fig. 4a-b, showing an increasing trend towards end-diastole for some heartbeats. Further work in a larger population is however necessary to verify these findings. A true load-independent measure of myocardial stiffness might be obtained by looking at the slope of the linear fit between diastolic wave speed and EDP during loading interventions (see Table 2), demonstrating a good reproducible fit for each pig. It still needs to be investigated whether (i) this slope alters during I/R injury, and (ii) a non-invasive loading change such as leg lifting can induce a change in wave speed that can be picked up. As an elevated pressure typically coincides with an elevated myocardial stiffness in cardiac disease, distinguishing a wave speed increase due to elevated pressure or stiffness might offer additional insights into pathophysiology.
Contractility
Systolic wave speed has been put forward as a measure for contractility in literature, because of its excellent correlation with ESP during isoproterenol stimulation 2 and with coronary perfusion pressure via the Gregg effect 26. The current study characterized contractility using two parameters, i.e. ESP and loading-independent PRSW. The results showed that systolic wave speed reflected the changes in contractility during the loading interventions (due to the Frank-Starling principle), as can be seen from the strong correlation with ESP in Fig. 6c. However, systolic wave speed increased significantly after ischemia injury (4.9 vs. 3.9 m/s; p = 0.01 in Fig. 5b), which does not correspond with the observed decline in contractility in terms of pressure-volume measures after the I/R injury (see Table 1). Also, no correlation was found between systolic wave speed and any measure of contractility during I/R injury (see Fig. 6c-d). In literature, contradictory results are found concerning systolic wave speed changes after I/R injury: decrease after 20 min ligation of the LAD 4 and (non-significant) increase after 1-2h ligation of the LAD 7.
In theory, an acute myocardial infarction is associated with an elevated myocardial stiffness and decreased contractility 27. We hypothesized that both mechanisms affect systolic wave speed in an opposing manner, and it is therefore unsure what the net resulting effect is. This study therefore investigated the ratio of the systolic and diastolic wave speed as potential index of contractility: the wave speed ratio moderately correlated with measures of contractility for loading interventions (R = 0.47; p = 0.02 for ESP and R = 0.54; p = 0.04 for PRSW in Fig. 7b-c), whereas a strong correlation was found for the stiffness interventions (R = 0.76; p < 0.001 for ESP and R = 0.73; p < 0.001 for PRSW in Fig. 7b-c). Furthermore, the equation of the linear regression line between wave speed ratio and PRSW was almost identical for loading and stiffness interventions, and resulted in \(\frac{SW{S}_{sys}}{SW{S}_{dia}}=0.028\bullet PRSW+1.4\) (R = 0.6; p < 0.001) when considering both type of interventions simultaneously in Fig. 8b. This demonstrates the same sensitivity of wave speed ratio to both types of interventions.
The sensitivity of the use of wave speed ratio as non-invasive index of contractility needs to be further investigated in a larger clinical study as the current study only explored a limited number of animals (n = 7) and the induced changes in contractility are larger than what is reported in cardiac disease, e.g. PRSW altered + 17.5% in hypertension and − 6.9% in heart failure with preserved ejection fraction 28.
Study limitations
It should be noted that SWE assesses systolic and diastolic ventricular properties in a local region (~ 3 cm) of the septal wall, whereas pressure-volume loop analysis provides functional measures of the left ventricular chamber as a whole. In this study, interventions were chosen such that they resulted in global alterations of left ventricular function or in local (septal) changes via LAD occlusion. Therefore, the local SWE measurements in the septal wall strongly correlated with the global pressure-volume measures. SWE measurements in different regions and walls of the heart can offer us insights into the regional differences of systolic and diastolic ventricular properties.
PRSW and EDPVR were not measured during preload decrease and afterload increase as this required insertion of a second balloon in the vena cava inferior and execution of two interventions at the same time. Also, right ventricular pressure was not measured even though this pressure also affects the resulting stresses and strains in the septal wall, and consequently wave speed. It is however unsure how much right ventricular pressure alters in case of an acute septal infarct.
As current study presented animal experiments, SWE measurements did not fulfill the FDA guidelines concerning acoustic safety for human exams. Acoustic safety can however be guaranteed through ECG gating or programming scanner off time after every SWE acquisition to reduce the spatial peak temporal average acoustic intensity. Concerning SWE data processing, diastolic and systolic wave speeds were now derived from a piecewise linear model, as suggested by Hollender et al. 13, but this model might need to be tuned to account for higher end-diastolic wave speeds (see Fig. 4).