# Magnetic resonance elastography in nonlinear viscoelastic materials under load

- 1.1k Downloads

## Abstract

Characterisation of soft tissue mechanical properties is a topic of increasing interest in translational and clinical research. Magnetic resonance elastography (MRE) has been used in this context to assess the mechanical properties of tissues in vivo noninvasively. Typically, these analyses rely on linear viscoelastic wave equations to assess material properties from measured wave dynamics. However, deformations that occur in some tissues (e.g. liver during respiration, heart during the cardiac cycle, or external compression during a breast exam) can yield *loading bias*, complicating the interpretation of tissue stiffness from MRE measurements. In this paper, it is shown how combined knowledge of a material’s rheology and loading state can be used to eliminate loading bias and enable interpretation of *intrinsic* (unloaded) stiffness properties. Equations are derived utilising perturbation theory and Cauchy’s equations of motion to demonstrate the impact of loading state on periodic steady-state wave behaviour in nonlinear viscoelastic materials. These equations demonstrate how loading bias yields apparent material stiffening, softening and anisotropy. MRE sensitivity to deformation is demonstrated in an experimental phantom, showing a loading bias of up to twofold. From an unbiased stiffness of \(4910.4 \pm 635.8\) Pa in unloaded state, the biased stiffness increases to 9767.5 \(\pm \,\)1949.9 Pa under a load of \(\approx \) 34% uniaxial compression. Integrating knowledge of phantom loading and rheology into a novel MRE reconstruction, it is shown that it is possible to characterise intrinsic material characteristics, eliminating the loading bias from MRE data. The framework introduced and demonstrated in phantoms illustrates a pathway that can be translated and applied to MRE in complex deforming tissues. This would contribute to a better assessment of material properties in soft tissues employing elastography.

## Keywords

Elastic waves Tissue mechanics Nonlinear mechanics Magnetic resonance elastography Biorheology## 1 Introduction

Diagnosis and therapy planning often depend on developing an accurate understanding of the state of a patient’s disease. Complex alterations occur in diseased tissue, depending on a range of factors such as changes to protein isoforms, alterations in protein density, extracellular matrix reorganisation, inflammation, etc. These modifications fundamentally impact tissue mechanical properties, triggering changes in stiffness, elasticity and viscosity (Yeh et al. 2002; Yin et al. 2009; Baiocchini et al. 2016). The ability to detect changes in tissue mechanical properties in vivo has the potential to significantly benefit clinical medicine, enabling a more thorough assessment of pathology as well as monitoring of treatment progression. This potential has lead to the development of a number of approaches for the noninvasive assessment of tissue mechanics.

Manual palpation is a classical approach of inspecting tissue for changes in elasticity in accessible organs (Weiss 2003; Nguyen et al. 2014). Advancements in medical imaging have led to the development of elastography—a modern quantitative technique for assessing the mechanical properties of internal organs. Elastography imaging is usually achieved using Ultrasound or Magnetic Resonance Imaging (MRI) (see Carey and Carey 2010 for a review), enabling the determination of tissue properties in vivo. Elastography imaging focuses on observing waves as they traverse through tissues, thus allowing quantification of strain response due to an excitation. In the case of magnetic resonance elastography (MRE), low-amplitude harmonic waves (10–1000, 30–80 Hz in vivo) are produced on the body and subsequently imaged in three-dimensions (Manduca et al. 2001; Fovargue et al. 2018b). Harmonic wave motion is then related to local tissue properties, in most cases, through linear viscoelastic wave equations (Glaser et al. 2012). MRE has been successfully employed for a range of tissues—including brain (Klatt et al. 2007; Green et al. 2008; Schregel et al. 2012), liver (Huwart et al. 2006, 2007, 2008; Klatt et al. 2007) and breast (Sinkus et al. 2000, 2005a, 2007; Houten et al. 2003)—showing contrast between diseased and healthy tissue as well as providing resolution of apparent tissue anisotropy (Qin et al. 2013). More recently, it has also been applied to heart (Elgeti et al. 2008; Kolipaka et al. 2009; Robert et al. 2009; Kolipaka et al. 2010; Couade et al. 2011; Kolipaka et al. 2012; Elgeti and Sack 2014), skeletal muscles (Green et al. 2012, 2013), and other soft tissues, reviewed in Manduca et al. (2001), Mariappan et al. (2010) and Fovargue et al. (2018b).

While interpreting stiffness measures in organs, such as the liver, has shown diagnostic value, a challenge yet to be addressed is the impact of macroscale deformations on wave propagation behaviour. Broadly, by looking at the 1D wave equation \(\partial ^2 u / \partial t^2 = c^2 \partial ^2 u / \partial x^2\), term \(c = \sqrt{T/\rho }\) depends on tension *T* and linear density \(\rho \) and is scaling the spatial second derivative of displacement *u*. The resultant force term on the right hand side balances the acceleration term on the left hand side, leading to the phase velocity increasing with higher tension. Elastic, hyperelastic and viscoelastic nonlinear materials would exhibit this effect. This is relevant in the context of tissues that undergo natural deformations (e.g: the heart muscle during the cardiac cycle (Asner et al. 2017), the liver during respiration (Kang et al. 2012), etc.) and tissues that undergo imposed deformations (e.g. breast during MRE scans Sinkus et al. 2005a, 2007). As many tissues are known to exhibit significant nonlinearity (Nash and Hunter 2000; Taber 2004; Liu et al. 2006; Holzapfel and Ogden 2009; Gao et al. 2010; Nordsletten et al. 2011b; Goenezen et al. 2012) under physiological conditions, both natural and imposed deformation can result in *loading bias*—an apparent stiffness that is a conglomerate measure of the intrinsic material properties and current kinematic state. For instance, the nonlinearity arising from compressive strain was quantified in ex vivo bovine liver samples using MRE measurements (Clarke et al. 2011). Hence, mixtures of patient-specific kinematics and intrinsic tissue properties are non-trivial and presumably result in a reduction in clinical specificity.

In this paper, the harmonic perturbation of Cauchy’s equations of motion for a general incompressible nonlinear viscoelastic material is examined, in order to understand the links between intrinsic properties and kinematics in MRE. From this analysis, a set of governing equations are derived, illustrating the general impact of large deformation on the motion of harmonic waves. Importantly, these equations enable the elimination of loading bias and therefore retrieval of intrinsic properties when large-scale deformation and constitutive behaviour are known. Theoretical test cases are presented, showing the ability of deformation to yield apparent increase, decrease and anisotropy in materials. In order to demonstrate the applicability of these perturbed equations, experiments were performed in polyvinyl alcohol (PVA) phantoms. Rheological tests were performed to characterise material behaviour, and loaded phantoms were imaged using MRE. Results show the capacity of these equations to correct for the complex dynamics of wave motion in deformed materials. To the authors’ knowledge, this is the first cross-validation study linking mechanical characterisation of a material in rheology with wave motion through the deformed material in MRE. The developments presented provide a key step towards improving the estimation of material properties in soft tissues, using elastography.

As a starting point, Sect. 2.1 presents perturbation theory employed within Cauchy’s equations of motion. The perturbed equations are analysed in the context of a general viscoelastic material under load, further leading to the derivation of apparent stiffness moduli influenced by large deformations and material constitutive law. For exemplification purposes, an idealised case of harmonic wave motion traversing a uniformly compressed Neo-Hookean material is presented. Experiments are employed to test the theoretical ground developed. In Sect. 2.2, PVA phantoms are tested in a rheological setup, to determine the material governing law. In Sect. 2.3, MRE data are acquired in PVA phantoms at different uniaxial compression levels. The knowledge on rheological behaviour and deformation is integrated into MRE data analysis, in order to undo the loading bias. The experimental results are presented in Sect. 3. Particularly, the estimation of the intrinsic stiffness of PVA is presented in Sect. 3.3. The theoretical framework presented in this study could be extended to in vivo MRE data analysis.

## 2 Materials and methods

### 2.1 Harmonic wave motion in deformed nonlinear viscoelastic materials

MRE relies on the analysis of small amplitude harmonic waves traversing through materials. As materials undergo deformation, wave behaviour is impacted (Thurston and Brugger 1964). One option to understand the relationship between large-scale deformation and small-scale wave behaviour is perturbation analysis. After covering notation 2.1.1, this section reviews perturbation results for Cauchy’s equations (Sect. 2.1.2) considering the particular case of a nonlinear viscoelastic material. Example cases are then highlighted (Sect. 2.1.3), illustrating the impact of load on the apparent stiffness characteristics stemming from MRE waves.

#### 2.1.1 Kinematics, harmonic wave motion and notation

*m*-dimensional function \( {\varvec{v}} : {\varOmega }_0\times [0, T] \rightarrow {\mathbb {R}}^m \) can be written as a function of the deformed domain \({\varOmega }(t)\), e.g. \( \varvec{\widetilde{v}} : {\varOmega }\rightarrow {\mathbb {R}}^m \), with the understanding that \(\varvec{\widetilde{v}}({\varvec{x}}, t) = {\varvec{v}}({\mathcal {L}}^{-1}({\varvec{x}},t),t)\). For ease of notation, the \(\widetilde{\cdot }\) is dropped where there is clear distinction.

The stresses resulting from material deformation can be separated into deviatoric and hydrostatic components. For the latter, the effect is quantified using the hydrostatic pressure *P*, which ensures some constraint on the volume or pressure-volume relation. The deviatoric components are, in turn, related to shape changes (or strains and strain history).

*J*:

^{1}

^{2}

*r*and

*i*, respectively.

#### 2.1.2 Perturbation of Cauchy’s equations of motion

*P*, and the right Cauchy Green tensor, \(\varvec{C}\).

*t*). We recall briefly that, for hyperelastic materials—such as those commonly applied in biomechanics—\( {\varvec{S}}_e \) is given as the derivative of a governing hyperelastic strain energy function, \( W = W_e(\varvec{C})\) (Bonet and Wood 2008), with respect to the right Cauchy Green tensor \(\varvec{C}\), i.e.

*P*. In the absence of deformation, under the assumption of an isotropic material defined by a standard law (e.g. a Neo-Hookean model for \({\varvec{S}}_e\) and \({\varvec{S}}_v\)), these stiffness moduli are simply given by \(\varvec{\mathcal {G}} ^\prime + i \varvec{\mathcal {G}} ^{\prime \prime } = G^* \varvec{{\mathcal {I}}}\), where \(G^*\) is the true complex shear modulus characterising the material and \({\mathcal {I}}_{ijkl} = \delta _{ik} \delta _{jl} \). The wave dynamics become symmetric, since \(\varvec{{\mathcal {I}}} : \nabla \varvec{u}_{{c}}= \nabla \varvec{u}_{{c}}+\nabla \varvec{u}_{{c}}^T\). However, under deformation, the components of \( \nabla \varvec{u}_{{c}}\) are scaled in a non-trivial way. By understanding the form of the scaling influencing the \( \varvec{\mathcal {G}} ^\prime \) and \( \varvec{\mathcal {G}} ^{\prime \prime } \) moduli, the bias introduced by deformation on the wave behaviour can be predicted. Furthermore, the deformation can be undone, that is, the moduli that reflect the undeformed state can be recovered.

#### 2.1.3 Loading bias of planar shear waves in pure compression

*J*, which are constant in space), and thus, the elastic contribution can be further simplified to

### 2.2 Nonlinear viscoelastic characterisation of PVA

Experiments done in controllable media, where shapes and deformations are simple, constitute the first step in evaluating the theory explained in the previous section. A first key component is using a material with a known rheological behaviour, as this governs the effective loading bias (e.g. Eq. 6). The following sections describe the fabrication process of PVA material and its testing in a rheological setup. A viscoelastic material law is also formulated. This law will then be later (Sect. 2.3.2) integrated with deformation and MRE data.

#### 2.2.1 PVA phantom preparation

During the freezing process, volumetric expansion occurred. Therefore, the phantoms’ top was cut, in order to achieve cuboid shapes. This led to differences in the phantoms’ height, which ranged between 32 and 42 mm. For a meaningful variance testing, 14 homogeneous phantoms were created, out of which 7 were used for rheological experiments and 7 for MRE experiments. However, one phantom was excluded in the rheological tests due to rupture. The time between phantom fabrication and testing in either rheology or MRE was within two weeks, to avoid potential material degradation.

#### 2.2.2 Rheological testing of PVA phantoms

The phantoms allow for suitable examination methods that can test the theory developed in Sect. 2.1. It was seen that, in order to account for the nonlinear complex stiffness moduli \( \varvec{\mathcal {G}} ^\prime \) and \( \varvec{\mathcal {G}} ^{\prime \prime }\) (Eq. 10), understanding of the nonlinear behaviour and frequency dependence of the material must be inferred. This can be done through rheological testing, which is used to characterise material behaviour by quantifying its response to different types of applied stress. As such, each of the six phantoms created for rheology experiments was tested in a Bose Electroforce 5500 test instrument. The setup, seen in Fig. 2, consisted of a supporting fixed platen and an adjustable upper platen, which was movable in the vertical direction. A load cell of capacity 225N measured the force required to ensure a user defined deformation.

For our aims, three types of tests were employed, with the purpose of assessing PVA’s nonlinear behaviour. The first test was designed to investigate the material’s response to dynamic micro-oscillations around a state of large static load. This design replicates closely the in vivo MRE scenario, where low-amplitude harmonic waves are propagating in an organ which may be subjected to large deformation. For reproducing the nonlinear behaviour, four different static loading states were investigated. Thus, the moving platen was programmed to compress the phantom by 2, 4, 6 and 8 mm, respectively (\(\sim \)5–20% uniaxial compression), and then to oscillate with an amplitude of 0.15 mm (\(\sim \)0.4%). This test will be referred to as the micro-oscillatory test.

The second test was designed to investigate the nonlinear large deformation response in time. As such, large oscillations were imposed around a state of large deformation: the platen compressed the phantom by 5.8 mm (\(\sim \)15% uniaxial compression) and then oscillations of 4 mm (\(\sim \)10%) amplitude were imposed. This test will be referred to as the macro-oscillatory test.

In both tests, oscillatory tests were carried under a frequency sweep, from 0.1 Hz to 10 Hz (the instrument’s limit): 0.1, 0.5, 1, 2, 5, 7.5 and 10 Hz. A clear depiction of the testing protocol can be seen in Fig. 2. This was done to investigate the frequency response of PVA, with the aim of extrapolating it to the higher frequencies used in MRE. In all tests, but clearly noticeable in the macro-oscillatory one, the displacements’ amplitude is decreasing with increasing frequency. This is probably due to the instrument’s limitation which, in the faster frequency regime, cannot reach the instructed displacements.

At the higher frequencies (\(\ge \, 5 \, \hbox {Hz}\)), it was observed that the moving platen gains acceleration and introduces a bias on the data. As such, calibration data were acquired, where the same protocol was followed, but with no sample in between the platens. Subtracting the calibration data from the phantom data eliminated the platen’s momentum bias.

A final stress relaxation test was conducted by subjecting the material to a constant large deformation, in order to capture the specific viscoelastic effect. Hence, the phantoms were held under a compression of 11 mm (\(\sim \)28.5%) for 5 min.

Overall, each phantom underwent six tests: four micro-oscillations at different loads (with seven different oscillatory states acquired), one macro-oscillation (with seven different oscillatory states acquired) and one relaxation. In between tests, the phantoms were left to rest in water for 15–20 min, since PVA is best stored in water (Surry et al. 2004), for hydration reasons. This resting time allowed for an efficient testing protocol of the phantoms, and its duration was sufficient to observe reproducibility.

#### 2.2.3 Constitutive modelling of PVA

#### 2.2.4 Rheology data analysis and model fitting

The viscoelastic model derived in the previous section (Eq. 19) needs to be tailored to the PVA material used in the rheological experiments by finding appropriate parameters \(C_1, \, C_2, \, \delta _1, \, \delta _2\) (Pa) and \(\alpha \).

*F*(

*t*) is the force reading, and

*A*(

*t*) is the area at time

*t*, which depends on the initial top area \(A_0\) and the ratio of initial cuboid height \(h_0\) to the compressed height

*h*(

*t*) (with \(h(t) = h_0 - d(t)\)).

*P*can also be determined at each time point by noticing that the traction in the \({\varvec{e}}_{1,2}\) directions is 0. Hence,

*P*counteracts the elastic and viscoelastic contributions in both the \({\varvec{e}}_{1,2}\) directions and can be used as such in computing the traction in the \({\varvec{e}}_3\) direction. Thus, the only unknowns in the traction model are parameters \(C_1, \, C_2, \, \delta _1, \, \delta _2, \, \alpha \). In broad terms, since \(\alpha \) is the only nonlinear parameter of the model, the parameterisation process is done by iterating over fixed values of \(\alpha \) and then solving a minimisation problem which yields a best fit of the remaining linear parameters. It is desired that all tests (micro-, macro-oscillations and relaxation) carry the same importance in the fitting process, and that the error is not dominated by certain tests (e.g. those done at higher compression levels, which employ larger tractions). As such, for each of the six tests (four micro-oscillation, one macro-oscillation and one relaxation), the traction data and model were standardised according to the maximum traction value in the test, i.e.

*k*. This mean value captures the compression level and, by considering locally every cycle in a test, captures the relaxation behaviour. Hence, the first part of the error deals with the sum, over all cycles, of the mean traction value difference between the data and the model, normalised by the data:

*k*is found as the difference between the traction value and the traction mean over the cycle, e.g.

Having defined the error to be minimised, a more detailed parameter fitting process can be described. The nonlinear parameter \(\alpha \) is iterated between 0.05 and 1 (close to the elastic and viscous limits, respectively), with an incremental step of 0.05. Since the PVA material used has a low viscoelastic response (and \(\alpha =0.05\) was observed to be the most suitable), the incremental step was refined to be 0.01 between 0.01 and 0.1. For each fixed \(\alpha \), the linear parameters \(C_1, \, C_2, \, \delta _1, \, \delta _2\) were searched by solving the least squares problem corresponding to error Eq. 20, using the inbuilt function *lsqnonlin* of MATLAB and Statistics Toolbox Release 2015a, The MathWorks, Inc., Natick, Massachusetts, USA. The nonlinear solver was used in order to allow for a positivity constraint on the parameters. For each phantom, a full set of parameters was determined.

To ensure parameter uniqueness, once the best \(\alpha \) value was determined, a new parameter search was performed. Thus, each individual linear parameter (\(C_1, \, C_2, \, \delta _1 \text { and } \delta _2\), respectively) was fit using the process described above, but while imposing the remaining three linear parameters to be 0 Pa. This was done in order to gauge the value of each model component and to identify potential parameter coupling.

### 2.3 Harmonic wave motion in deformed PVA

#### 2.3.1 MRE experimental setup

When designing the MRE experiment, it was important to obtain shear waves that would propagate through the whole body of the phantom, while maintaining a fixed position of the phantom. The setup used is illustrated in Fig. 3. A transducer indenting the phantom transmits small amplitude compressional waves. Since the PVA material is aqueous, it could easily slide on a flat surface. For this reason, the phantom’s back side was in contact with the support, ensuring no penetration, hence converting the compressional waves into shear waves. A large uniaxial deformation could be obtained by simply compressing the phantom with a smooth top plate, as seen in Fig. 3. Due to the sliding feature of the PVA material, the phantoms did not bulge under the uniaxial compression just described.

Using this setup inside the MR scanner, coronal wave data were recorded using MRE in undeformed and deformed configurations. Each phantom was scanned in the reference state (uncompressed), first deformation state consisting of uniaxial compression of \(\sim \)5 mm (\(\sim \)14%) and second deformation state consisting of uniaxial compression of \(\sim \)12 mm (\(\sim \)35%).

MR acquisition consisted of two primary scans. First, a geometric scan was acquired providing 55 cross-sectional slices with a field of view (FOV) of \(96 \times 96 \, \hbox {mm}\) (resolution of \(96 \times 96\) pixels and slice thickness of 1 mm). Scan parameters were adjusted to enhance signal of the PVA material (TE/TR= 120.00 /5000.00 ms).

The second scan was a MRE eXpresso Gradient Echo sequence (Garteiser et al. 2013) providing 10 cross-sectional slices with a FOV of \(96 \times 96 \, \hbox {mm}\) (resolution \(96 \times 96\) pixels and slice thickness of 2.0 mm). Driver frequency and motion encoding gradients were set, by turn, at 120, 130 and 140 Hz, and TE/TR at 6.91 /151.82 ms. The sequence encoded each of the three displacement component direction—\( \varvec{e}_1 \), \( \varvec{e}_2 \), \( \varvec{e}_3 \) and a reference non-motion encoded image removing background phase shifts due to the MRI gradients. Examples of the geometric and wave images can be seen in Fig. 4, for the undeformed and deformed configurations.

#### 2.3.2 MRE data analysis

The geometric data acquired were used to provide detailed quantitative information of the geometrical configuration of the material in reference and deformed states. Assuming that the cuboid phantoms were compressed uniformly, then the deformation gradient can be written as in Eq. 12, where \(\lambda \) is the compression defined as the ratio of the compressed and uncompressed heights. Hence, due to the design of the MRE experiment, this form of the deformation gradient can be employed with the stiffness moduli given by Eq. 10.

This is usually separated into the (real) storage modulus \(G^{\prime }\) and (imaginary) loss modulus \(G^{\prime \prime }\) (where \(G^* = G^{\prime } + i G^{\prime \prime }\)).

As explained in Sect. 2.1.2, the form of \(\varvec{\mathcal {G}} ^*\) presented in Eq. 21 holds true in the absence of deformation and under an isotropic standard material law, and the wave gradient contribution becomes symmetric, as \(\varvec{{\mathcal {I}}} : \nabla \varvec{u}_{{c}}= \nabla \varvec{u}_{{c}}+ \nabla \varvec{u}_{{c}}^T\). Hence, this method is suitable for reconstructing the stiffness modulus when no deformation is employed, but will lead to a biased result under deformation. Throughout the rest of the paper, this method will be referred to as uncorrected reconstruction (UR).

As previously mentioned, within the UR the wave gradient is contracted with the fourth-order identity tensor scaled by the complex stiffness modulus, as \((G^* \varvec{\mathcal {I}} ):\nabla \varvec{u}_{{c}}\). By contrast, the CR allows the wave to be contracted with two distinct fourth-order tensors, defined by the user. One result is to be integrated with the real part and one with the imaginary part of the complex stiffness modulus, as \({M} (\varvec{\mathcal {G}} ^\prime : \nabla \varvec{u}_{{c}}) + iN (\varvec{\mathcal {G}} ^{\prime \prime } : \nabla \varvec{u}_{{c}})\), where \(M+iN\) is to be found. This is a novel way of analysing MRE data and is of utmost importance, since we saw, in Eq. 10, that the real and imaginary moduli \(\varvec{\mathcal {G}} ^{\prime }\) and \(\varvec{\mathcal {G}} ^{\prime \prime }\) are non-trivial and change differently with deformation. These changes depend on metrics derived from the known deformation gradient \({\varvec{F}}\) (\({\varvec{B}}\), \(I_{{\varvec{C}}}\), etc.), and, for our PVA material, are scaled by the parameters determining the viscoelastic Eq. 19. The specific form can be found in Appendix 2, which is applicable to the PVA material used here. Thus, assuming that the viscoelastic model describes the data well, then \(\varvec{\mathcal {G}} ^{\prime }\) and \(\varvec{\mathcal {G}} ^{\prime \prime }\) already incorporate the true real and imaginary stiffness moduli. Hence, it is predicted that *M* and *N* will be reconstructed as unity.

*M*and

*N*were reconstructed as closely as possible to unity for each pixel, over all pixels. The nonlinear parameter \(\alpha \) was fixed to be the one indicated by rheology tests. The set of parameters thus obtained from the reference and second compression states was tested against the first compression state, for prediction value. The error was measured as the pixel-wise difference between the reconstructed values

*M*and unity in uncompressed and second compression case, normalised by the number of pixels, as

*n*is the total number of pixels considered and subscript

*k*iterates over those pixels.

## 3 Results and discussion

The first essential step in the experimental work was tailoring the viscoelastic law 19 to the PVA phantoms used in the rheology testing. The results of this process are presented in Sect. 3.1 and help us understand the rheological behaviour of the PVA material used, which directly influences the loading bias. The deformation bias is shown in Sect. 3.2, by investigating uniaxial compression in the MRE experimental setup. In Sect. 3.3 it is shown that, by incorporating the information on rheology and deformation into the CR, the intrinsic stiffness of the PVA material can be retrieved from the loaded states of the phantoms.

### 3.1 Nonlinear viscoelastic model for PVA

The aim of the modelling process was to fit the viscoelastic model 19 to the PVA data acquired in the rheological setup. Hence, for each of the six phantoms used in the rheology experiment, the parameters were fit accounting for all oscillatory and relaxation tests simultaneously. It will be seen that the parameters cannot be uniquely identified, which stems from the fact that the PVA has a small viscoelastic response, as it will be explained shortly. As such, a model with a reduced number of parameters is sought, which ensures unique parameter estimation.

Best fit parameters for each PVA phantom; the error, computed as per Eq. 20, can be seen in the first column (in brackets)

Phantom | \(C_1\) (Pa) | \(C_2\) (Pa) | \(\delta _1\) (Pa) | \(\delta _2\) (Pa) | \(\alpha \) |
---|---|---|---|---|---|

p1 (3.44%) | 1545.66 | 0.18 | 195.76 | 149.40 | 0.06 |

p2 (5.69%) | 1683.89 | 11.55 | 0.47 | 275.13 | 0.05 |

p3 (1.50%) | 1606.03 | 4.50 | 1457.86 | 193.03 | 0.04 |

p4 (1.71%) | 2491.84 | 57.71 | 628.60 | 79.40 | 0.09 |

p5 (1.01%) | 647.09 | 174.31 | 2259.66 | 2.00 | 0.03 |

p6 (6.69%) | 1746.13 | 0.04 | 0.00 | 272.97 | 0.05 |

Mean | 1620.11 | 41.38 | 757.06 | 161.99 | 0.05 |

SD | 537.34 | 62.71 | 839.25 | 98.87 | 0.019 |

Best fit parameter for each PVA phantom; the error, computed as per Eq. 20, can be seen below each parameter; each parameter was fit considering the remaining three parameters to be 0

Phantom | \(C_1\) (Pa) | \(C_2\) (Pa) | \(\delta _1\) (Pa) | \(\delta _2\) (Pa) | \(\alpha \) |
---|---|---|---|---|---|

p1 | 1835.42 | 685.33 | 2175.25 | 813.40 | 0.03 |

err (%) | 11.69 | 53.97 | 9.80 | 58.12 | |

p2 | 1933.73 | 884.40 | 2318.04 | 1031.81 | 0.03 |

err (%) | 17.78 | 52.34 | 15.18 | 59.42 | |

p3 | 3015.56 | 1190.96 | 3568.47 | 1433.38 | 0.03 |

err (%) | 9.29 | 52.93 | 7.31 | 56.10 | |

p4 | 3036.63 | 1072.38 | 3585.49 | 1258.29 | 0.03 |

err (%) | 7.89 | 57.27 | 6.24 | 60.80 | |

p5 | 2864.40 | 1132.68 | 3374.09 | 1346.91 | 0.03 |

err (%) | 8.70 | 53.30 | 8.70 | 56.92 | |

p6 | 1934.22 | 890.65 | 2363.49 | 922.69 | 0.03 |

err (%) | 22.35 | 52.31 | 15.39 | 65.47 | |

Mean | 2436.66 | 976.06 | 2897.47 | 1134.41 | NA |

SD | 590.75 | 189.85 | 677.24 | 248.31 | NA |

When investigating \(C_2\) , the nonlinearity of the \({\varvec{S}}_{e2}\) component becomes obvious—a linear increase in the compression level (2 mm between the micro-oscillatory tests) leads to a nonlinear response in the traction force. Hence, this parameter alone provides an unsuitable fit to the data; however, it adds value to the full model by being able to adjust for the increased amplitude in the four micro-oscillation tests. As before, the viscoelastic counterpart \(\delta _2\) captures the relaxation behaviour. Table 2 presents the best fit parameter for each PVA phantom. By investigating Fig. 7, it is expected that the \(C_1\) and \(\delta _1\) parameters provide a better fit, and hence yield smaller errors than the \(C_2\) and \(\delta _2\) parameters.

Best fit parameters for each phantom, restricting \(C_1, \delta _2 = 0\) Pa; the error, computed as per Eq. 20, can be seen in the first column, in brackets

Phantom | \(C_2\) (Pa) | \(\delta _1\) (Pa) | \(\alpha \) |
---|---|---|---|

p1 (3.95%) | 118.94 | 1923.03 | 0.03 |

p2 (6.54%) | 233.16 | 1900.25 | 0.03 |

p3 (1.78%) | 169.10 | 3220.10 | 0.03 |

p4 (2.32%) | 117.86 | 3143.31 | 0.02 |

p5 (1.28%) | 176.42 | 2854.77 | 0.02 |

p6 (8.74%) | 216.15 | 1973.74 | 0.03 |

Linear | Linear | Nonlin | |

Mean | 171.94 | 2502.53 | 0.03 |

SD | 43.71 | 581.34 | 0.005 |

The new set of parameters obtained for the 3-parameter model can be seen in Table 3. It is notable that the error increase is very small; thus, the new parametrisation does not significantly deteriorate the quality of the fit. An example of the parameterised 3-parameter model fit to the rheology data can be seen in Fig. 6 (right). The error plot of the 3-parameter model can be seen in Fig. 5 (right). Each phantom indicated a best fit for a very small \(\alpha \) (0.02 or 0.03). At low \(\alpha \) values, the errors are generally small, since the PVA is only slightly viscoelastic. However, at high \(\alpha \) values, there is a sharp increase in the minimal error, which eventually plateaus. This happens when the contribution of the viscoelastic parameter becomes insignificant (\(\delta _1\) becomes very small, to counteract the effect of \(\alpha \)) and the only remaining parameter is thus \(C_2\). In this case, as seen in Table 2 as well, the errors rise up to \(\sim \)30–35%. In the low \(\alpha \) regime, employing the 3-parameter model increases the errors only slightly compared to the 5-parameter model, which is preferred due to reduced complexity.

### 3.2 Loading bias in harmonic wave motion under pure compression

In this study, the PVA phantom material is described by the moderately complex viscoelastic Eq. 24. Harmonic waves were designed to probe the PVA, as shown in Fig. 4 (columns 3–5).

MRE minimisation parameters fixing \(\alpha =0.03\); the error, computed as per Eq. 23, can be seen in the first column, in brackets

Phantom | \(C_2\) (Pa) | \(\delta _1\) (Pa) | \(\alpha \) |
---|---|---|---|

p7 (15.85%) | 93.68 | 2119.92 | 0.03 |

p8 (15.86%) | 83.15 | 2085.28 | 0.03 |

p9 (14.80%) | 88.36 | 1878.05 | 0.03 |

p10 (16.64%) | 94.83 | 2126.26 | 0.03 |

p11 (17.08%) | 103.43 | 2235.11 | 0.03 |

p12 (18.27%) | 106.27 | 1945.07 | 0.03 |

p13 (11.15%) | 246.54 | 1848.72 | 0.03 |

Linear | Nonlin | ||

Mean | 120.43 | 2019.75 | NA |

SD | 56.96 | 139.60 | NA |

### 3.3 Intrinsic stiffness estimation

By understanding how loading biases stiffness estimation, it is possible to estimate intrinsic stiffness values when using CR. This requires knowledge of the deformation employed and the material law. Since this study deals with pure compression, the deformation was estimated using Eq. 12. Rheological testing of the PVA led to a 3-parameter material law (Eq. 24). It was seen, in Sect. 3.1, that there is variation in estimating the linear parameters \(C_2\) and \(\delta _1\), whereas the nonlinear parameter \(\alpha \) is stable around 0.02–0.03. The optimisation performed in rheology was not repeated for MRE, as this was observed to lead to a parameter coupling between \(\alpha \) and \(\delta _1\). Thus, \(\alpha \) was fixed at 0.03, as indicated by the rheological tests, and \(C_2\), \(\delta _1\) were estimated as described in the data analysis Sect. 2.3.2. The resulting parameters are presented in Table 4. It is notable that the \(C_2\) parameter is generally lower in MRE (Table 4) than in rheology (Table 3), with the exception of phantom 13. This could be related to the degree of relative compression achieved. In rheology, the maximum relative compression is 28.5%. In MRE, however, relative compressions reach up to 35%, with the exception of phantom 13 (only 18%), as seen in Fig. 9. At lower deformation levels, the viscoelastic component is less employed; thus, the purely elastic part becomes more dominant. As \(C_2\) is scaling the elastic component, its estimate is higher when less uniaxial compression is achieved. Within each MRE phantom, experimental imperfections lead to variability between pixels. Looking at the apparent stiffness mapsin Fig. 8, excluding the boundary area (as shown in column 2), a small degree of heterogeneity is observed, which accentuates under higher compressions. One reason behind this is speculated to lie in the crystallisation process during the F-T cycles of PVA. Assuming that the PVA freezes first at the mould edges and then in the centre, this could affect the local structures formed. Another reason could be that the top side of the phantoms is not perfectly flat, as it was cut manually. As such, the compression could have not been ideally uniform. Other arguments could revolve around the compression that the piston is exerting on the phantom, but an analysis in this sense is intricate and beyond the purpose of this study. Regardless of the underlying reasons, the small local variations in each phantom lead to the errors presented in Table 4, as slightly different pixels are attempted to be fit to the same measure. Although the errors are higher than in rheology (Table 3), the error metrics are not directly comparable. In rheology, a transient behaviour is investigated, that is, a data point is followed in time (Eq. 20), whereas in MRE a large spatial sample size is examined—all the pixels across several slices of each phantom, in uncompressed and second compressed state (Eq. 23).

The parameters obtained in Table 4 were sought by employing the CR on each phantom in uncompressed and second compression states, as previously described. Subsequently, they were used directly in reconstructing the corrected maps *M* for the first compression case, correspondingly for each phantom. The corrected maps *M* are illustrated, in a phantom, in Fig. 8 (column 3). Since the parameters are fit to best describe the data in uncompressed and second compressed states, it was indeed expected that the *M* map’s average for these cases is close to unity. Additionally, the corrected map for the first compression case, whose average also lies close to unity, demonstrates the prediction value of the parameters.

The quantified average of map *M* for all phantoms in all deformation states (uncompressed, first compression and second compression) is shown in Fig. 9 (right). Same as for the apparent stiffness \(G^{\prime }\), only a ROI was considered inside each phantom, to avoid boundary effects. Depending on each phantom’s height, a variable compression \(\lambda \) describes the first or second compression. For instance, the data points on the graph corresponding to \(\lambda \) between 0.75 and 0.8 depict the first compression for taller phantoms and second compression for shorter phantoms. It is observed that the standard deviations are sizeable (10–19%) and tend to be higher at the larger compression levels. In the stiffness measurements, this observation was attributed to the apparent anisotropy effect and waves’ local probing directions. In the corrected maps *M*, the apparent anisotropy effect is undone by incorporating the knowledge on the deformation gradient, given by Eq. 12. Since the variance does not decrease in the *M* maps compared to \(G^{\prime }\) maps, it suggests that the predicted apparent anisotropy in compression is not the main cause of the increasing standard deviations. Instead, it could be due to experimental imperfections which are difficult to control. For instance, the slight unevenness of the phantoms’ top or the compression induced by the piston indentation are unavoidable, yet for simplicity their effects are not accounted for in the deformation gradient, which employs an idealised form. Despite these, in all but one cases, the average of corrected map *M* lies within 10% of the ideal value of 1. This indicates that the model, parameterised with the values presented in Table 4, captures the bias induced by loading on the wave motion. This shows that the parameters found can describe the material at different deformation states and are suitable for estimating the intrinsic stiffness of the PVA material.

Integrating rheology and MRE is a difficult problem to tackle. While rheology can be employed to directly measure a material’s stress-strain response, MRE relies on investigating the wave propagation behaviour in order to estimate material properties. Here, we tried to integrate the two methods by validating a viscoelastic model using rheological data and then using it to derive the \(\varvec{\mathcal {G}} ^{\prime }\) and \(\varvec{\mathcal {G}} ^{\prime \prime }\) moduli to be used within CR of MRE data. The model was shown to capture the loading bias, thus proving the successful integration of the methods. Furthermore, model parameterisation was done separately for rheology and MRE, yet it is notable that the standard deviations around the mean of the \(C_2\) and \(\delta _1\) parameters are overlapping across the two experimental methods. This, together with the undoing of the loading bias, demonstrates the improvement that the newly developed CR brings in analysing MRE data.

### 3.4 Extension and impact in tissue

The theoretical framework presented in Sect. 2.1.2 showed that the stiffness moduli are directly related to the material law and deformation. The work done in phantoms supports the theoretical concept. It was shown that, by incorporating the knowledge on material behaviour and deformation into the MRE reconstruction, the intrinsic stiffness can be estimated. This approach could be extended and applied in tissues. As a first step, a biomechanical model describing the tissue could be inferred and further adapted from animal rheological work. Additionally, by employing registration tools on anatomical images of an organ in undeformed and deformed states, deformations can be quantified. Together, a better characterisation of specific tissue rheology and large deformation recordings using MRI could enable proper interpretation of MRE data.

The most immediate extension of our approach could be breast MRE. This imaging technique relies on the breast being fixed in between two plates, which leads to tissue compression. In the literature, the loading bias was observed in cases where the breast was fixed tightly, compared to cases where the breast was fixed more loosely (Sinkus et al. 2005a), but was not accounted for. Undoing this bias using subject specific deformation would help in correcting the observed stiffness for both healthy and diseased areas. For instance, provided that the risk of breast cancer is associated with tissue stiffness (Boyd et al. 2014), accounting for the loading bias would ensure that the risk assessment is not underestimated or overestimated. Clinically, this could improve the prognostic value and disease management.

Another extension of our study could be in the liver, which is subjected to motion due to respiration. While this results mainly in translational motion, parts of the liver can be strained (Kang et al. 2012). This can be advantageous, since some liver pathologies were shown to become more apparent, compared to healthy tissue, under large strains (in ex vivo) (Yeh et al. 2002). Measuring stiffness in strained tissue could increase diagnostic accuracy, but would require techniques to eliminate loading bias.

The heart is a particularly complex organ which presents structural anisotropy, has dynamic stiffness, and undergoes large motion during the cardiac cycle. The latter constitutes one of the many challenges in assessing the unbiased stiffness of the myocardium (Kolipaka et al. 2010; Glaser et al. 2012), since the effects of deformation non-trivially interfere with the intrinsic tissue properties. The framework presented in this study would constitute a step forward in assessing the tissue characteristics, by eliminating the deformation bias.

In order to follow this paper’s proposed pathway of retrieving intrinsic material parameters from MRE data, one would have to consider Eq. 9 in the material frame as a starting point. Since the moduli \(\varvec{\mathcal {G}} ^{\prime }\) and \(\varvec{\mathcal {G}} ^{\prime \prime }\) depend on constitutive behaviour and deformation (e.g. Eq. 10), then knowledge on the PK2 tensor and deformation gradient is necessary. The material behaviour can be modelled by considering a constitutive equation for the strain energy function *W*, like in Eq. 17, from which tensor \({\varvec{S}}\) can be derived, or by proposing an adaption for \({\varvec{S}}\), e.g. Eq. 19. A formulation for the deformation gradient \({\varvec{F}}\) is required, like the one presented in Eq. 12, which needs to describe the unperturbed macro-deformation that the body is undergoing. From these, a form of the moduli \(\varvec{\mathcal {G}} ^{\prime }\) and \(\varvec{\mathcal {G}} ^{\prime \prime }\) can be inferred, which can be employed within CR (Eq. 22) to eliminate the loading bias.

### 3.5 Study limitations

A limitation encountered during the analysis process of the data presented in this study was the estimation of the viscous modulus from MRE data. The rheological tests performed on PVA phantoms identified a very low \(\alpha \), indicating that there is little viscous response of the material. This is a consequence of choosing a highly elastic material, with low wave attenuation. As such, the characteristics of the viscous component were difficult to identify under MRE testing, leading to an overestimation of the reconstructed viscous modulus. This effect was already identified with the employed MRE reconstruction (Fovargue et al. 2018a), where the overestimation of the viscous modulus was observed in a purely elastic phantom, particularly with increased noise. However, this drawback is predicted to disappear when using in vivo data, as tissue has a stronger viscous component. Nevertheless, in this phantom study, the bias due to the spectrum of stiffness prevented the characterisation of the full complex modulus, thus shifting the focus only on the real stiffness modulus.

This study provides the means through which a material’s intrinsic characteristics can be retrieved, despite it experiencing loading. However, a reasonable quantification of intrinsic material parameters relies on a good estimation of the deformation state and on an appropriate constitutive material law. In more complex scenarios, like in vivo measurements, these prerequisites are more challenging to attain. The deformation metrics would presumably need to rely on precise image registration techniques. As medical image registration is a greatly developed field, identification of suitable methods is surely attainable. However, the material constitutive law could be difficult to model, particularly in tissues which exhibit anisotropy. Nonetheless, much work has been done in this sense, as tissue models have been developed in many forms—polynomial, exponential, logarithmic, power laws, etc., and can be further improved and adapted. Undeniably, a model with increased complexity would lead to convoluted \(\varvec{\mathcal {G}} ^{\prime }\) and \(\varvec{\mathcal {G}} ^{\prime \prime }\) moduli, yet they can be attained numerically (by contrast with the analytic form presented here, for PVA, in Appendix 2). Hence, despite an increased computational cost, the use of a complex model would not be hindered.

## 4 Conclusions

This paper presents a framework for integrating knowledge on material constitutive behaviour and deformation into wave dynamics in order to retrieve intrinsic material stiffness properties. First off, a theoretical foundation was established by perturbing Cauchy’s equations of motion and transcribing the resulting equations into material frame. This determined a form of the stiffness moduli which was seen to depend on the material model and deformation metrics. To support the theory, experiments on PVA phantoms were employed. The PVA material was tested in a rheological setup and its examined behaviour was modelled using a viscoelastic law. The same material was subjected to MRE experiments, where a known deformation (uniaxial compression) was employed. It was seen that, under the designed testing conditions, using standard MRE analysis, the material appeared to be stiffer with increasing compression. Thus, the known material model and deformation were integrated into the stiffness moduli and subsequently into the MRE reconstruction. By doing this, it was shown that intrinsic material stiffness parameters can be estimated, thus undoing the loading bias observed using standard analysis. Hence, this study provides a framework demonstrated to work in phantoms, which can be adapted and applied to MRE in more complex instances.

## Footnotes

- 1.
Einstein summation over repetitive indices is assumed.

- 2.
Only dimensionality

*m*was used as this case is relevant to this work, particularly \(m=3\). - 3.Here, we denote bythe directional derivative of$$\begin{aligned} D_{{\varvec{x}}}[f][{\varvec{u}} ] = \lim _{\varepsilon \rightarrow 0} \frac{f({\varvec{x}} + \varepsilon {\varvec{u}} ) - f({\varvec{x}})}{\varepsilon } \end{aligned}$$
*f*at a point \({\varvec{x}}\) in the direction of increment \({\varvec{u}}\).

## Notes

### Compliance with ethical standards

### Conflict of interest

The authors declare that they have no conflict of interest.

## References

- Anseth KS, Bowman CN, Brannon-Peppas L (1996) Mechanical properties of hydrogels and their experimental determination. Biomaterials 17(17):1647–1657Google Scholar
- Asner L, Hadjicharalambous M, Chabiniok R, Peressutti D, Sammut E, Wong J, Carr-White G, Razavi R, King AP, Smith N, Lee J, Nordsletten D (2017) Patient-specific modeling for left ventricular mechanics using data-driven boundary energies. Comput Methods Appl Mech Eng 314:269–295MathSciNetGoogle Scholar
- Baiocchini A, Montaldo C, Conigliaro A, Grimaldi A, Correani V, Mura F, Ciccosanti F, Rotiroti N, Brenna A, Montalbano M, D’Offizi G, Capobianchi MR, Alessandro R, Piacentini M, Schininà ME, Maras B, Del Nonno F, Tripodi M, Mancone C (2016) Extracellular matrix molecular remodeling in human liver fibrosis evolution. PLoS One 11(3):e0151736Google Scholar
- Banks HT, Hu S, Kenz ZR (2011) A brief review of elasticity and viscoelasticity for solids. Adv Appl Math Mech 3(1):1–51MathSciNetGoogle Scholar
- Bilston LE, Liu Z, Phan-Thien N (2001) Large strain behaviour of brain tissue in shear: some experimental data and differential constitutive model. Biorheology 38(4):335–345Google Scholar
- Bonet J, Wood RD (2008) Nonlinear continuum mechanics for finite element analysisGoogle Scholar
- Boyd NF, Li Q, Melnichouk O, Huszti E, Martin LJ, Gunasekara A, Mawdsley G, Yaffe MJ, Minkin S (2014) Evidence that breast tissue stiffness is associated with risk of breast cancer. PLoS One 9(7):e100937Google Scholar
- Caputo M (1967) Linear models of dissipation whose Q is almost frequency independent-II. Geophys J Int 13(5):529–539Google Scholar
- Carey E, Carey WD (2010) Noninvasive tests for liver disease, fibrosis, and cirrhosis: Is liver biopsy obsolete? Cleve Clin J Med 77:519–527Google Scholar
- Chui C, Kobayashi E, Chen X, Hisada T, Sakuma I (2004) Combined compression and elongation experiments and non-linear modelling of liver tissue for surgical simulation. Med Biol Eng Comput 42(6):787–798Google Scholar
- Clarke EC, Cheng S, Green M, Sinkus R, Bilston LE (2011) Using static preload with magnetic resonance elastography to estimate large strain viscoelastic properties of bovine liver. J Biomech 44(13):2461–2465Google Scholar
- Couade M, Pernot M, Messas E, Bel A, Ba M, Hagege A, Fink M, Tanter M (2011) In vivo quantitative mapping of myocardial stiffening and transmural anisotropy during the cardiac cycle. IEEE Trans Med Imaging 30(2):295–305Google Scholar
- Elgeti T, Rump J, Hamhaber U, Papazoglou S, Hamm B, Braun J, Sack I (2008) Cardiac magnetic resonance elastography. Initial results. Invest Radiol 43(11):762–772Google Scholar
- Elgeti T (2014) Sack I (2014) Magnetic resonance elastography of the heart. Curr Cardiovasc Imaging Rep 7:9247Google Scholar
- Fovargue D, Kozerke S, Sinkus R, Nordsletten D (2018a) Robust MR elastography stiffness quantification using a localized divergence free finite element reconstruction. Med Image Anal 44:126–142Google Scholar
- Fovargue D, Nordsletten D, Sinkus R (2018b) Stiffness reconstruction methods for MR elastography. NMR Biomed 18:e3935Google Scholar
- Gamage TPB, Rajagopal V, Ehrgott M, Nash MP, Nielsen PMF (2011) Identification of mechanical properties of heterogeneous soft bodies using gravity loading. Int J Numer Methods Biomed Eng 27(3):391–407zbMATHGoogle Scholar
- Gao Z, Lister K, Desai JP (2010) Constitutive modeling of liver tissue: experiment and theory. Ann Biomed Eng 38(2):505–516Google Scholar
- Garteiser P, Sahebjavaher RS, Ter Beek LC, Salcudean S, Vilgrain V, Van Beers BE, Sinkus R (2013) Rapid acquisition of multifrequency, multislice and multidirectional MR elastography data with a fractionally encoded gradient echo sequence. NMR Biomed 26(10):1326–1335Google Scholar
- Gasser TC, Ogden RW, Holzapfel GA (2006) Hyperelastic modelling of arterial layers with distributed collagen fibre orientations. J R Soc Interface 3(6):15–35Google Scholar
- Gennisson JL, Rénier M, Catheline S, Barrière C, Bercoff J, Tanter M, Fink M (2007) Acoustoelasticity in soft solids: assessment of the nonlinear shear modulus with the acoustic radiation force. J Acoust Soc Am 122(6):3211–3219Google Scholar
- Glaser KJ, Manduca A, Ehman RL (2012) Review of MR elastography applications and recent developments. J Magn Reson Imaging 36(4):757–774Google Scholar
- Goenezen S, Dord JF, Sink Z, Barbone PE, Jiang J, Hall TJ, Oberai AA (2012) Linear and nonlinear elastic modulus imaging: an application to breast cancer diagnosis. IEEE Trans Med Imaging 31(8):1628–1637Google Scholar
- Golub GH, Van Loan CF (1996) Matrix computations. Johns Hopkins University Press, BaltimorezbMATHGoogle Scholar
- Graff KF (1991) Wave motion in elastic solids. Dover Publications, New YorkzbMATHGoogle Scholar
- Green MA, Bilston LE, Sinkus R (2008) In vivo brain viscoelastic properties measured by magnetic resonance elastography. NMR Biomed 21(7):755–764Google Scholar
- Green MA, Sinkus R, Gandevia SC, Herbert RD, Bilston LE (2012) Measuring changes in muscle stiffness after eccentric exercise using elastography. NMR Biomed 25(6):852–858Google Scholar
- Green MA, Geng G, Qin E, Sinkus R, Gandevia SC, Bilston LE (2013) Measuring anisotropic muscle stiffness properties using elastography. NMR Biomed 26(11):1387–1394Google Scholar
- Gurtin ME, Fried E, Anand L (2010) The mechanics and thermodynamics of continua. Cambridge University Press, CambridgeGoogle Scholar
- Hadjicharalambous M, Lee J, Smith NP, Nordsletten DA (2014) A displacement-based finite element formulation for incompressible and nearly-incompressible cardiac mechanics. Comput Methods Appl Mech Eng 274:213–236MathSciNetzbMATHGoogle Scholar
- Hariton I, DeBotton G, Gasser TC, Holzapfel GA (2007) Stress-driven collagen fiber remodeling in arterial walls. Biomech Model Mechanobiol 6(3):163–175Google Scholar
- Holzapfel G (2000) Nonlinear solid mechanics: a continuum approach for engineering, 1st edn. Wiley, LondonzbMATHGoogle Scholar
- Holzapfel GA, Ogden RW (2009) Constitutive modelling of passive myocardium: a structurally based framework for material characterization. Philos Trans R Soc A Math Phys Eng Sci 367(1902):3445–3475MathSciNetzbMATHGoogle Scholar
- Huwart L, Peeters F, Sinkus R, Annet L, Salameh N, ter Beek LC, Horsmans Y, Van Beers BE (2006) Liver fibrosis: non-invasive assessment with MR elastography. NMR Biomed 19(2):173–179Google Scholar
- Huwart L, Sempoux C, Salameh N, Jamart J, Annet L, Sinkus R, Peeters F, ter Beek LC, Horsmans Y, Van Beers BE (2007) Liver fibrosis: noninvasive assessment with mr elastography versus aspartate aminotransferase-to-platelet ratio index. Radiology 245(2):458–466Google Scholar
- Huwart L, Sempoux C, Vicaut E, Salameh N, Annet L, Danse E, Peeters F, ter Beek LC, Rahier J, Sinkus R, Horsmans Y, Van Beers BE (2008) Magnetic resonance elastography for the noninvasive staging of liver fibrosis. Gastroenterology 135(1):32–40Google Scholar
- Kang N, Lee MW, Rhee T (2012) Simulating liver deformation during respiration using sparse local features. IEEE Comput Graph Appl 32(5):29–38Google Scholar
- King DM, Moran CM, McNamara JD, Fagan AJ, Browne JE (2011) Development of a vessel-mimicking material for use in anatomically realistic Doppler flow phantoms. Ultrasound Med Biol 37(5):813–826Google Scholar
- Kiss MZ, Varghese T, Hall TJ (2004) Viscoelastic characterization of in vitro canine tissue. Phys Med Biol 49(18):4207–4218Google Scholar
- Klatt D, Hamhaber U, Asbach P, Braun J, Sack I (2007) Noninvasive assessment of the rheological behavior of human organs using multifrequency MR elastography: a study of brain and liver viscoelasticity. Phys Med Biol 52(24):7281–7294Google Scholar
- Kolipaka A, McGee KP, Araoz PA, Glaser KJ, Manduca A, Romano AJ, Ehman RL (2009) MR elastography as a method for the assessment of myocardial stiffness: comparison with an established pressure-volume model in a left ventricular model of the heart. Magn Reson Med 62(1):135–140Google Scholar
- Kolipaka A, Araoz PA, McGee KP, Manduca A, Ehman RL (2010) Magnetic resonance elastography as a method for the assessment of effective myocardial stiffness throughout the cardiac cycle. Magn Reson Med 64(3):862–870Google Scholar
- Kolipaka A, Aggarwal SR, McGee KP, Anavekar N, Manduca A, Ehman RL, Araoz PA (2012) Magnetic resonance elastography as a method to estimate myocardial contractility. J Magn Reson Imaging 36(1):120–127Google Scholar
- Liu Z, Bilston L (2000) On the viscoelastic character of liver tissue: experiments and modelling of the linear behaviour. Biorheology 37(3):191–201Google Scholar
- Liu Y, Wang G, Sun LZ (2006) Anisotropic elastography for local passive properties and active contractility of myocardium from dynamic heart imaging sequence. Int J Biomed Imaging 2006:45957Google Scholar
- Malvern LE (1969) Introduction to the mechanics of a continuous medium. Prentice-Hall, Englewood CliffsGoogle Scholar
- Manduca A, Oliphant TE, Dresner MA, Mahowald JL, Kruse SA, Amromin E, Felmlee JP, Greenleaf JF, Ehman RL (2001) Magnetic resonance elastography: non-invasive mapping of tissue elasticity. Med Image Anal 5(4):237–254Google Scholar
- Mariappan YK, Glaser KJ, Richard LE (2010) Magnetic resonance elastography: a review. Clin Anat 23(5):497–511Google Scholar
- McCormick M, Nordsletten D, Kay D, Smith N (2013) Simulating left ventricular fluid-solid mechanics through the cardiac cycle under LVAD support. J Comput Phys 244:80–96MathSciNetzbMATHGoogle Scholar
- McCormick M, Nordsletten D, Lamata P, Smith NP (2014) Computational analysis of the importance of flow synchrony for cardiac ventricular assist devices. Comput Biol Med 49:83–94Google Scholar
- Nash MP, Hunter PJ (2000) Computational mechanics of the heart. J Elast 61(1/3):113–141zbMATHGoogle Scholar
- Nguyen MM, Zhou S, Robert JL, Shamdasani V, Xie H (2014) Development of oil-in-gelatin phantoms for viscoelasticity measurement in ultrasound shear wave elastography. Ultrasound Med Biol 40(1):168–176Google Scholar
- Nicolle S (2015) Comportement viscoélastique non linéaire des organes pleins de l ’ abdomen. Congrès Français de MécaniqueGoogle Scholar
- Nicolle S, Vezin P, Palierne JF (2010) A strain-hardening bi-power law for the nonlinear behaviour of biological soft tissues. J Biomech 43(5):927–932Google Scholar
- Nordsletten D, McCormick M, Kilner PJ, Hunter P, Kay D, Smith NP (2011a) Fluid-solid coupling for the investigation of diastolic and systolic human left ventricular function. Int J Numer Method Biomed Eng 27(7):1017–1039MathSciNetzbMATHGoogle Scholar
- Nordsletten D, Niederer S, Nash M, Hunter P, Smith N (2011b) Coupling multi-physics models to cardiac mechanics. Prog Biophys Mol Biol 104(1–3):77–88Google Scholar
- Ogden RW (1997) Non-linear elastic deformations. Dover Publications, New YorkGoogle Scholar
- Pazos V, Mongrain R, Tardif JC (2009) Polyvinyl alcohol cryogel: optimizing the parameters of cryogenic treatment using hyperelastic models. J Mech Behav Biomed Mater 2(5):542–549Google Scholar
- Qin EC, Sinkus R, Geng G, Cheng S, Green M, Rae CD, Bilston LE (2013) Combining MR elastography and diffusion tensor imaging for the assessment of anisotropic mechanical properties: a phantom study. J Magn Reson Imaging 37(1):217–226Google Scholar
- Rajagopal V, Nielsen PM, Nash MP (2010) Modeling breast biomechanics for multi-modal image analysis-successes and challenges. Wiley Interdiscip Rev Syst Biol Med 2(3):293–304Google Scholar
- Reynolds HM, Puthran J, Doyle A, Jones W, Nielsen PMF, Nash MP, Rajagopal V (2011) Mapping breast cancer between clinical X-ray and MR images. Computational biomechanics for medicine, pp 81–90. Springer, New YorkGoogle Scholar
- Robert B, Sinkus R, Gennisson J, Fink M (2009) Application of dense-MR-elastography to the human heart. Magn Reson Med 62(5):1155–1163Google Scholar
- Schregel K, Wuerfel E, Garteiser P, Gemeinhardt I, Prozorovski T, Aktas O, Merz H, Petersen D, Wuerfel J, Sinkus R (2012) Demyelination reduces brain parenchymal stiffness quantified in vivo by magnetic resonance elastography. Proc Natl Acad Sci 109(17):6650–6655Google Scholar
- Sinkus R, Lorenzen J, Schrader D, Lorenzen M, Dargatz M, Holz D (2000) High-resolution tensor MR elastography for breast tumour detection. Phys Med Biol 45(6):1649–1664Google Scholar
- Sinkus R, Tanter M, Catheline S, Lorenzen J, Kuhl C, Sondermann E, Fink M (2005a) Imaging anisotropic and viscous properties of breast tissue by magnetic resonance-elastography. Magn Reson Med 53(2):372–387Google Scholar
- Sinkus R, Tanter M, Xydeas T, Catheline S, Bercoff J, Fink M (2005b) Viscoelastic shear properties of in vivo breast lesions measured by MR elastography. Magn Reson Imaging 23(2):159–165Google Scholar
- Sinkus R, Siegmann K, Xydeas T, Tanter M, Claussen C, Fink M (2007) MR elastography of breast lesions: understanding the solid/liquid duality can improve the specificity of contrast-enhanced MR mammography. Magn Reson Med 58(6):1135–1144Google Scholar
- Surry KJM, Austin HJB, Fenster A, Peters TM (2004) Poly(vinyl alcohol) cryogel phantoms for use in ultrasound and MR imaging. Phys Med Biol 49(24):5529–5546Google Scholar
- Taber LA (2004) Nonlinear theory of elasticity. World Scientific, SingaporezbMATHGoogle Scholar
- Thurston RN, Brugger K (1964) Third-order elastic constants and the velocity of small amplitude elastic waves in homogeneously stressed media. Phys Rev 133(6A):A1604–A1610zbMATHGoogle Scholar
- Van Houten EH, Weaver JB, Miga MI, Kennedy FE, Paulsen KD (2000) Elasticity reconstruction from experimental MR displacement data: initial experience with an overlapping subzone finite element inversion process. Med Phys 27(1):101–107Google Scholar
- Van Houten EE, Doyley MM, Kennedy FE, Weaver JB, Paulsen KD (2003) Initial in vivo experience with steady-state subzone-based MR elastography of the human breast. J Magn Reson Imaging 17(1):72–85Google Scholar
- Wang CC, Truesdell C (1973) Introduction to Rational Elasticity (Mechanics of Continua). SpringerGoogle Scholar
- Wang VY, Lam HI, Ennis DB, Cowan BR, Young AA, Nash MP (2009) Modelling passive diastolic mechanics with quantitative MRI of cardiac structure and function. Med Image Anal 13(5):773–784Google Scholar
- Weiss NS (2003) Breast cancer mortality in relation to clinical breast examination and breast self-examination. Breast J 9(Suppl 2):S86–S89Google Scholar
- Xia W, Piras D, Heijblom M, Steenbergen W, van Leeuwen TG, Manohar S (2011) Poly(vinyl alcohol) gels as photoacoustic breast phantoms revisited. J Biomed Opt 16(7):075002Google Scholar
- Yeh WC, Li PC, Jeng YM, Hsu HC, Kuo PL, Li ML, Yang PM, Lee PH (2002) Elastic modulus measurements of human liver and correlation with pathology. Ultrasound Med Biol 28(4):467–474Google Scholar
- Yin M, Chen J, Glaser KJ, Talwalkar JA, Ehman RL (2009) Abdominal magnetic resonance elastography. Top Magn Reson Imaging 20(2):79–87Google Scholar

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.