Image-Based Estimation of Left Ventricular Myocardial Stiffness

Abstract

Elevation in left ventricular (LV) myocardial stiffness is a key remodeling-mediated change that underlies the development and progression of heart failure (HF). Despite the potential diagnostic value of quantifying this deterministic change, there is a lack of enabling techniques that can be readily incorporated into current clinical practice. To address this unmet clinical need, we propose a simple protocol for processing routine echocardiographic imaging data to provide an index of left ventricular myocardial stiffness, with protocol specification for patients at risk for heart failure with preserved ejection fraction. We demonstrate our protocol in both a preclinical and clinical setting, with representative findings that suggest sensitivity and translational feasibility of obtained estimates.

1 Introduction

Heart failure (HF) continues to be a leading cause of death and disability, and despite advances in HF management, the HF burden continues to rise at an alarming rate [1,2]. Using available Medicare data, hospital readmission due to HF have estimated costs exceeding $50 billion by next year and approaching $100 billion by 2030—costs that are certain to impact medical resources and reach an unsustainable level [3,4]. One major barrier to preventing HF progression is that while patients may present with clinical symptoms of HF, the underlying pathophysiology can be quite different. An ever increasing and particularly challenging pathophysiological classification is HF with preserved ejection fraction (HFpEF) [5–7]. Unlike HF with reduced ejection fraction, wherein compromised left ventricular (LV) pump function can be easily detected, HFpEF has a subtle and insidious progression that conventional imaging approaches such as ultrasound (echocardiography) are not able to identify at early stages [8]. The underlying pathophysiology of HFpEF is that of primarily diastolic dysfunction that manifests when increases in LV wall thickness and LV myocardial stiffness together cause a progressive increase in LV chamber stiffness. Increased LV chamber stiffness in turn leads to elevated LV filling pressures and ultimately the development of HF signs and symptoms. Currently, echocardiographic methods cannot directly measure a key response variable in HFpEF development and progression, namely, LV myocardial stiffness, and thus this critical functional milestone is not available for clinical decision making. We proposed to address this unmet medical need by developing a protocol which takes advantage of speckle tracking echocardiography (STE) to establish a straightforward approach to provide an index of LV myocardial stiffness, where the requisite data processing can be readily integrated into a conventional echocardiography workflow/workstation.

Speckle tracking echocardiography provides a noninvasive tool to quantify regional LV myocardial strain based on relative changes in a segment length, which are defined with respect to acoustic markers (speckles) that are contained within a specified myocardial region and tracked over the cardiac cycle [9]. In the LV short-axis echocardiographic view, regional LV myocardial strain/strain rate can be obtained in the radial and circumferential directions using correspondingly oriented segments; in the LV long-axis view, regional radial and longitudinal LV myocardial strain/strain rate can be similarly measured. Thus, STE-based strain measures are linearized approximations of the normal strain components that would, in addition to shear strain components, be appropriate for a complete description of LV myocardial finite deformation in the framework of continuum solid mechanics [10]. While these approximations limit the applicability of STE-based LV myocardial strain in fundamental mechanics, they are clinically accessible and have demonstrated utility for HF (primarily HF with reduced ejection fraction) diagnosis and prognosis in both preclinical and clinical settings [11–17].

Speckle tracking echocardiography-based LV myocardial strains depend on LV mechanical loading and geometry, as well as the mechanical properties of the LV myocardium. The LV myocardium exhibits layer-specific passive and active mechanical properties, with complex behavior that includes significant nonlinearity and anisotropy, contains residual strains, and therefore requires a rigorous application of continuum mechanics for complete constitutive modeling [18–22]. Thus, consideration of STE-based LV myocardial strain alone does not facilitate constitutive modeling of the LV myocardium, and along with inherent complex behavior has to date largely precluded the assessment of LV myocardial mechanical properties in routine clinical settings. We posit that the lack of translational methods to estimate LV myocardial mechanical properties represents a critical knowledge gap, as an increase in LV myocardial stiffness due to maladaptive LV remodeling underlies various modes of HF [17,23–26] and provides a direct or indirect therapeutic target for diverse cardiac interventions [27]. We seek to address this knowledge gap with a simple protocol to compute indices of LV myocardial stiffness via coupling of STE-based LV myocardial strain measures and approximations of regional LV myocardial stress.

2 Methods

2.1 Study Groups.

We first demonstrate and validate our protocol using representative echocardiographic data obtained from previous preclinical (porcine) studies of LV remodeling [25]; specifically, we consider an animal exhibiting maladaptive LV remodeling due to progressive left ventricular pressure overload (LVPO) that led to increased LV chamber stiffness, elevated LV myocardial collagen content, and preserved EF, as well as two age-matched referent control animals (one each for protocol demonstration and validation). Our focus on pressure-mediated maladaptive LV remodeling is clinically motivated by its central role in the development of HFpEF, wherein a primary output of the remodeling process is an increase in LV myocardial stiffness [23–26]. Moreover, because LVPO induces concentric and hypertrophic remodeling [28], we expect a reasonable correspondence between actual LV geometry and a simple geometrical model that facilitates estimation of LV myocardial stress (as detailed below).

Next, to support the translational feasibility, we demonstrate our protocol on routine clinical images obtained from both a non-HFpEF and an HFpEF patient. Both patients have diabetes managed with oral agents and had been completely revascularized for multivessel coronary heart disease. The patient with HFpEF underwent surgical replacement of his aortic valve 2 years prior to the echocardiogram and had persistent Class 2 HF symptoms managed with 1 mg bumetanide daily. Measured LV end-diastolic pressure was 18 mmHg, H2FPEF score was 4 (BMI-34.2, Hypertension, E/e′ > 9), and brain natriuretic peptide was 82 pg/mL. The patient without HFpEF was asymptomatic, LV end-diastolic pressure was 6 mmHg, and H2FPEF score was 0.

2.2 Mechanical Model: Applicability, Assumptions, and Governing Equations.

Our protocol is specified for assessment of LV myocardial stiffness indices as manifest in the diastolic phase of the cardiac cycle. Therefore, obtained indices are most relevant to the passive LV myocardial behavior (and thus reflect the effects of cardiac fibrosis prototypical of HFpEF) as opposed to LV myocardial contractility (which is more relevant to systolic HF modalities).

We focus on the LV short-axis view at the level of the papillary muscle and consider the midwall LV myocardial deformation from end-systole (ES, reference configuration) to the end-diastole (ED, deformed configuration). Regional myocardial strains $(\epsilon )$ are thus defined via the general relation

$$\epsilon =\frac{l_{\mathrm{ED}}-l_{\mathrm{ES}}}{l_{\mathrm{ES}}}$$ (1)

where $l_{\mathrm{ED}}$ and $l_{\mathrm{ES}}$ are the lengths of a regionally contained segment at ED and ES, respectively, with segment and associated strain orientations in either the radial ${(\epsilon }_r)$ or circumferential ${(\epsilon }_{\theta })$ direction. Note that although some STE analytical software considers ED as the reference configuration and computes a segmental strain at ES (i.e., GE EchoPAC™), a simple transformation of obtained measures can be applied to yield the strain as defined in Eq. (1). For example, if a typical STE strain output $\left({\epsilon }^{*}\right)$ is reported as a fractional length change with ${\epsilon }^{*}=\frac{l_{\mathrm{ES}}-l_{\mathrm{ED}}}{l_{\mathrm{ED}}}$, then $\epsilon =\frac{-{\epsilon }^{*}}{(1+{\epsilon }^{*})}$.

In the LV short-axis view, the kinematics and geometry of diastolic LV myocardial deformation in the considered cases can be reasonably modeled as a thick-walled cylinder that is inflated by an internal pressure. To specify the boundary value problem, we assume that in the reference configuration (ES) the LV chamber pressure is zero; in the deformed configuration (ED), the LV chamber pressure is equal to the pulmonary capillary wedge pressure. We ignore the presence of residual strains in the LV myocardium, and thus assume the traction-free (zero-pressure) reference configuration is also a zero-stress/zero-strain configuration. Out-of-plane myocardial deformation (i.e., twisting) and associated shear strains are not accounted for in our protocol, which is an inherent limitation of the obtained indices for LV myocardial stiffness.

To estimate mean regional wall stress $(\sigma )$ in the deformed configuration, we apply the universal solutions for a uniformly inflated thick-walled circular cylinder at mechanical equilibrium. Within a specified LV myocardial region (for which STE-based strains have been measured), we compute the associated mean LV myocardial stresses in the radial ${(\sigma }_r)$ and circumferential $\left({\sigma }_{\theta }\right)$ directions as

$${\sigma }_r=\frac{-P}{2}\hspace{1em}\mathrm{a}\mathrm{nd}\hspace{1em}{\sigma }_{\theta }=\frac{P{r}_i}{t}$$ (2)

where $P$ is the LV chamber pressure; $r_i$ is the deformed inner radius of the LV; $t$ is the deformed LV myocardial wall thickness; and $r_i$ and $t$ are referred to a specific LV myocardial region. Note that these expressions for mean stresses are derived for a circular cylindrical cross section with a traction-free outer surface in a state of axisymmetric deformation, and their use despite potential regional variations in geometry (due to deviations from a circular cross section, wherein the circle center is identified as the point-of-intersection of regional-defined vectors that are normal to the endocardial trace at ES) and outer surface boundary conditions (i.e., free wall versus septal wall) yields only approximate values. While ${\sigma }_r$ only depends on the mechanical load, regional variations in LV myocardial geometry are explicitly accounted for in computed values for ${\sigma }_{\theta }$.

We assume that the slope of the linearized regional stress/strain relation when going from the reference to the deformed configuration can be considered as an index of regional diastolic LV myocardial stiffness ( $K_M)$. Here again, our protocol does not consider an established feature of passive LV myocardial behavior, namely, significant nonlinearity, and thus provides a nominal stiffness as opposed to a constitutive material property. Thus, the regional LV myocardial stiffness indices in the radial ( $K_{M,r})$ and circumferential $\left(K_{M,\theta }\right)$ directions are computed via ED stress/strain ratios, namely,

$$K_{M,r}=\frac{{\sigma }_r}{{\epsilon }_r}\hspace{1em}\mathrm{and}{\hspace{1em}\hspace{0.17em}K}_{M,\theta }=\frac{{\sigma }_{\theta }}{{\epsilon }_{\theta }}$$ (3)

Given the kinematics of diastolic deformation, $K_{M,r}$ and $K_{M,\theta }$ provide indices of LV myocardial stiffness under compression and tension, respectively.

3 Results and Discussion

3.1 Preclinical Examples.

All animals were treated and cared for in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (Eighth Edition. Washington, DC: 2011), and all protocols were approved by the University of South Carolina School of Medicine and WJB Dorn VA Institutional Animal Care and Use Committee. To minimize pain and distress during echocardiographic studies, animals were administered oral diazepam (200 mg) 1 h prior to the imaging procedure and supplemental midazolam (0.5–0.6 mg/kg) intramuscularly at the time of the echocardiogram.

Our protocol for estimation of LV myocardial stiffness utilizes short-axis LV echocardiograms (GE Vivid E9 with XDclear Ultrasound System: M5S (1.5–4.6 Hz) transducer probe; GE, Boston, MA) from representative, age-matched animals extracted from previous studies on LV remodeling (referent control and LVPO, as described above). First, we used commercial echocardiographic analysis software (EchoPAC™) and established methods to measure regional LV wall geometry (wall thickness and inner radius) at ES and ED (Fig. 1), as well as conventional Doppler methods to measure pulmonary capillary wedge pressure [29]. In all cases, regional wall geometry in both the reference (ES) and deformed (ED) configurations is relatively uniform (as indicated by the standard deviation of the regional average of each measure), supporting the employed geometric/kinematic model of diastolic LV deformation (the uniform inflation of a thick-walled cylindrical tube).

Fig. 1.

Left ventricle geometry; reference and deformed configurations. Specification and quantification of the reference and deformed LV configurations used for estimation of LV myocardial stiffness. STE-based definition of LV myocardial regions in the short-axis echocardiographic view (referent control animal; AS-anterior septal; A-anterior; L-lateral; P-posterior; I-inferior; S-septal; referent control) and corresponding measures of LV chamber radius and myocardial wall thickness for the reference (a)–(c) and deformed (d)–(f) LV configurations. Black columns (▮) refer to referent control animal; white columns (◻) refer to LVPO animal; and regional averages and standard deviations are included for all geometric measures.

In line with our protocol, regional radial and circumferential LV myocardial strains (Eq. (1)), mean LV myocardial stress (Eq. (2)), and corresponding estimates of LV myocardial stiffness (Eq. (3)) were computed for a referent control and LVPO animal (Figs. 2(a)– 2(f)). Comparison between obtained values of $K_{M,r}$ (Fig. 2(c)) and $K_{M,\theta }$ (Fig. 2(f)) in these representative animals suggests our protocol can detect the regional and directional variance in LV myocardial stiffness, as well as qualitatively capture the expected increase in LV myocardial stiffness with LVPO. Moreover, the generally order-of-magnitude difference between $K_{M,r}$ and $K_{M,\theta }$ reflects the expected anisotropic behavior of the LV myocardium, with the comparatively high values of $K_{M,\theta }$ due to the characteristic behavior of the primary load-bearing wall constituent (collagen) under tension.

Fig. 2.

Left ventricle myocardial strain, stress, and stiffness. LV myocardial strains, LV myocardial stress, and LV myocardial stiffness in the radial (a)–(c) and circumferential (d)–(f) directions. Black columns (▮) refer to referent control animal; white columns (◻) refer to LVPO animal; and regional averages and standard deviations are included for all measures with regional variation. (g)Representative compressive stress–strain response from ex vivo uniaxial testing of referent control LV myocardium. Solid circle (●) represents relevant data for isometric comparison with STE-based LV myocardial stiffness estimates. (h) Correlation between regional LV myocardial stiffness estimates derived via ex vivo uniaxial mechanical testing and STE-based analysis (referent control animal; AS-anterior septal; A-anterior; L-lateral; P-posterior; I-inferior; S-septal).

To validate STE-based estimation of LV myocardial stiffness, we compared computed $K_{M,r}$ values to analogous values derived from regionally and directionally matched ex vivo mechanical testing of LV myocardium in a second reference control animal. Following established mechanical testing protocols [30], LV myocardial samples were obtained the day following echocardiographic studies on a referent control animal, prepared as cylindrical test elements (∼5 mm diameter; 3 mm length), and subjected to an unconfined radial compression test (ramped uniaxial displacement; displacement rate of 0.05 mm/s; total displacement of ∼1.5 mm) using a mechanical testing apparatus appropriated for soft tissue analyses (Bose^® Biodynamic Test Instrument, Minnetonka, MN). Resultant force and displacement data were continuously recorded (data scan rate of 200/s) by the system load cell and software package (WinTest^® Software, Minnetonka, MN) and transformed to yield a compressive stress–strain relation (Fig. 2(g)). The stress–strain coordinate corresponding to the STE-derived ${\epsilon }_r$ was identified for each LV myocardial region (solid data point in Fig. 2(g)) and subsequently used to calculate radial LV myocardial stiffness per Eq. (3), which again does not reflect a constitutive mechanical property of the LV myocardium that would require data from (at minimum) bi-axial mechanical testing [31]. Isometric comparison of $K_{M,r}$ resulting from these techniques suggests a systematic overestimation of LV myocardial stiffness with STE when compared to ex vivo testing ( $K_{M,r}$ of 2.9±0.26 kPa versus 1.5±0.57 kPa, respectively), but excellent correlation (R = 0.85) among regional values supports the diagnostic potential of our protocol (Fig. 2(h)). Moreover, general agreement between STE-derived $K_{M,r}$ values obtained in different referent control animals used in this study (Figs. 2(c) and 2(h)—x-axis, with regional values of 2.2±0.27 kPa and 2.9±0.26 kPa, respectively) supports protocol repeatability.

3.2 Clinical Examples.

Routine short-axis echocardiographic imaging (Fig. 3(a)) and STE-based measures of LV myocardial strains (Fig. 3(b)) were obtained for a non-HFpEF and HFpEF patient (as described above). Based on available image-quality and resultant speckle tracking, mechanical/geometrical measurements were restricted to the posterior LV myocardial region for the purpose of demonstrating the translational feasibility of our protocol. STE-based estimation of LV myocardial stiffness shows notable differences between non-HFpEF and HFpEF patients, with approximate 16- and 7-fold differences observed in the radial (Fig. 3(c)) and circumferential (Fig. 3(d)) directions, respectively. Notably, these differences were greater than analogous strain comparisons between patients, wherein recorded radial and circumferential LV myocardial strains exhibited ∼6-fold and 2.5-fold differences, respectively.

Fig. 3.

Clinical translation of LV myocardial stiffness estimation. (a) Representative short-axis echocardiographic view obtained from routine clinical assessment, with the posterior LV myocardium identified as the region of interest (ROI) for strain measure. (b) STE-based radial (left) and circumferential (right) LV myocardial strains, as automatically defined by STE software (i.e., end-diastolic to end-systolic strain). STE-based posterior region LV myocardial stiffness in the (c) radial and (d) circumferential directions computed for a non-HFpEF and HFpEF patient.

3.3 Limitations.

Our protocol extends current clinical use of STE to provide potentially useful indices for HFpEF therapy, but underlying assumptions and limitations should be considered in the interpretation of obtained values. Key simplifications that enable this protocol include (i) the imposed boundary value problem/kinematics to model LV geometry/deformation (uniform inflation of a thick-walled cylinder, as opposed to more detailed ellipsoidal models or direct geometric representation modeled via a finite element approach that are comparatively more difficult to translate) [17,32], (ii) the use of STE-derived LV myocardial strains (as opposed to nonlinear strains), (iii) no account of residual strains, and (iv) the use of linear stress/strain relations to compute indices of LV myocardial stiffness (as opposed to constitutive properties that model nonlinear, anisotropic, heterogeneous behavior in the framework of continuum solid mechanics).

4 Conclusion

We present a protocol for processing of STE-derived LV myocardial strain data to generate estimates of LV myocardial stiffness. Maladaptive LV remodeling is a canonical process in HFpEF, wherein increased LV myocardial stiffness contributes to disease progression and insomuch represents a potentially valuable response variable to improve clinical decision making. This protocol demonstration details a novel methodology for estimating LV myocardial stiffness in both a preclinical and clinical setting; subsequent studies will be performed to (i) further validate obtained LV myocardial stiffness estimates (via comparison with ex vivo measures of LV myocardial stiffness in a preclinical model) and (ii) further establish the clinical utility of estimating LV myocardial stiffness (via correlation between LV myocardial stiffness, LV chamber stiffness, and HFpEF development/progression in patients).

Funding Data

• National Heart, Lung, and Blood Institute, National Institutes of Health (Grant Nos. R01-HL159620, R01HL130972-01A1, and R01HL5949; Funder ID: 10.13039/100000050).

• Merit Award from the Veterans Health Administration (Grant Nos. BX000168-10A1 and BX005320; Funder ID: 10.13039/100006812).

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.