You are on page 1of 18

Physics in Medicine & Biology

PAPER

An analytical model of full-field displacement and strain induced by


amplitude-modulated focused ultrasound in harmonic motion imaging
To cite this article: Matthew D J McGarry et al 2021 Phys. Med. Biol. 66 075017

View the article online for updates and enhancements.

This content was downloaded from IP address 157.193.14.89 on 31/07/2021 at 23:22


Phys. Med. Biol. 66 (2021) 075017 https://doi.org/10.1088/1361-6560/abddd1

PAPER

An analytical model of full-field displacement and strain induced by


RECEIVED
23 July 2019
amplitude-modulated focused ultrasound in harmonic motion
REVISED
30 December 2020
imaging
ACCEPTED FOR PUBLICATION
20 January 2021
PUBLISHED
Matthew D J McGarry1,2,4 , Adriaan Campo1,4 , Thomas Payen1, Yang Han1 and Elisa E Konofagou1,3,∗
6 April 2021 1
Department of Biomedical Engineering, Columbia University, NY 10023, New York, United States of America
2
Thayer School of Engineering, Dartmouth College, Hanover NH 03755, United States of America
3
Department of Radiology, Columbia University, NY 10032, New York, United States of America
4
Both authors contributed equally to this manuscript.

Author to whom any correspondence should be addressed.
E-mail: ek2191@columbia.edu

Keywords: HMI, harmonic motion imaging, ultrasound, elastography, finite element model, FEM, analytic model

Abstract
The majority of disease processes involves changes in the micro-structure of the affected tissue, which
can translate to changes in the mechanical properties of the corresponding tissue. Harmonic motion
imaging (HMI) is an elasticity imaging technique that allows the study of the mechanical parameters of
tissue by detecting the tissue response by a harmonic motion field, which is generated by oscillatory
acoustic radiation force. HMI has been demonstrated in tumor detection and characterization as well as
monitoring of ablation procedures. In this study, an analytical HMI model is demonstrated and
compared with a finite element model (FEM), allowing rapid and accurate computation of the
displacement, strain, and shear wave velocity (SWV) at any location in a homogenous linear elastic
material. Average absolute differences between the analytical model and the FEM were respectively 1.2%
for the displacements and 0.5% for the strains for 41 940 force voxels at 0.22 s per displacement
evaluation. A convergence study showed that the average difference could be further decreased to 1.0%
and 0.15% for the displacements and strains, respectively, if force resolution is increased. SWV fields, as
calculated with the FEM and the analytical model, have regional differences in velocities up to 0.57 m s−1
with an average absolute difference of 0.11±0.07 m s−1, primarily due to imperfections in the non-
reflecting FEM boundary conditions. The apparent SWV differed from the commonly used plane-wave
approximation by up to 1.2 m s−1 due to near and intermediate field effects. Maximum displacement
amplitudes for a model with an inclusion stabilize within 10% of the homogenous model at an inclusion
radius of 10 mm while the maximum strain reacts faster, stabilizing at an inclusion radius of 3 mm. In
conclusion, an analytical model for HMI stiffness estimation is presented in this paper. The analytical
model has advantages over FEM as the full-field displacements do not need to be calculated to evaluate
the model at a single measurement point. This advantage, together with the computational speed, makes
the analytical model useful for real-time imaging applications. However, the analytical model was found
to have restrictive assumptions on tissue homogeneity and infinite dimensions, while the FEM
approaches were shown adaptable to variable geometry and non-homogenous properties.

1. Introduction

Microscale structural changes in tissue from disease processes manifest as changes in macroscale mechanical
properties of tissue, which are the basis for physicians using palpation as a primary diagnostic tool for centuries.
The field of elastography aims to create images of the mechanical properties of tissue for diagnosis and
monitoring of disease processes. In order to probe the mechanical properties of tissue, some sort of
displacement must be induced.

© 2021 Institute of Physics and Engineering in Medicine


Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Several such displacement methods have been used in elastography. External quasi-static compression is
common in, e.g. ultrasound strain imaging (Ophir et al 1991). Intrinsic displacements resulting from
physiologic processes such as respiratory and cardiac cycles can also provide information (Konofagou et al 2002,
Weaver et al 2012, Catheline et al 2013). Additionally, transient (Sandrin et al 2003) or harmonic (Muthupillai
et al 1995) external vibrations can be applied to probe the mechanical properties of internal tissues.
The acoustic radiation force (ARF) resulting from focused ultrasound (FUS) has been used to produce
localized internal displacements. Over the past two decades, several ARF-based ultrasound elasticity imaging
techniques have been developed to estimate tissue stiffness and thus detect tumors using various forms of tissue
perturbation to detect stiffer masses.
ARF results when propagating ultrasound waves transfer their momentum to the surrounding medium
through absorption and reflection. The magnitude of the force per unit volume FV (kg s−2 m−2) is given by:
2a abs ITA
FV = , (1)
c
where αabs (Np m−1) is the absorption of the medium, ITA (W m−2) is the temporal average intensity of the
ultrasound field at a given location, and c (m s−1) is the sound speed in the medium (Torr 1984). The resulting
deformation caused by ARF is described by Callé et al (2005).
ARF-based elasticity imaging methods remotely interrogate tissue mechanical properties by generating
impulsive radiation force (i.e. acoustic radiation force impulse imaging Nightingale et al 2002, Lizzi et al 2003),
spatially modulated ultrasound radiation force (McAleavey et al 2009), shear wave elasticity imaging (Sarvazyan
et al 1998), supersonic shear imaging (Bercoff et al 2004), shear wave spectroscopy (Deffieux et al 2009), and
comb-push ultrasound shear elastography (Song et al 2012). Shear wave-based techniques estimate the speed of
the transient shear wave, as the velocity of the shear wave (SWV) propagating away from the focus gives a
quantitative estimate of the shear modulus (Widman et al 2015). For a planar shear wave, the theoretical
SWVtheory is related to the Poisson’s ratio ν, Young’s modulus E and the density ρ (or related to the shear
modulus m and the density ρ) as follows:
E m
SWVtheory = = . (2)
2 (1 + v ) r r
Other strategies include quasi-static approaches (Ophir et al 1991) and or harmonic radiation force such as
vibro-acoustography (Fatemi and Greenleaf 1999, Fatemi et al 2002, Alizad et al 2012), shear wave dispersion
ultrasound vibrometry (Shigao et al 2009), and finally, harmonic motion imaging (HMI) (Konofagou and
Hynynen 2003).
HMI, the focus of this study, amplitude modulates the FUS source at a fixed frequency of 50 Hz to produce
an internally generated harmonic motion field (Konofagou and Hynynen 2003). HMI is capable of monitoring
the displacement in seamless synchronization with the application of radiation force, with applications to
monitoring thermal ablation and tissue viscoelasticity evaluation (Konofagou and Hynynen 2003, Maleke et al
2006, Curiel et al 2009a, 2009b, Chen et al 2015, Han et al 2016, Suomi et al 2016). Like other methods that
employ harmonic radiation forces to induce an oscillatory tissue motion, HMI allows efficient reduction of
noise generated by physiologic motion in displacement estimation because the frequencies of those noises
generated by physiologic motion, such as breathing, blood flow, and digestive tract movements (typically in the
range of 0.0–3.3 Hz Redmond and Hegge 1985), are distinct from the frequency of the oscillatory tissue motion
induced. The feasibility of HMI in tissue stiffness detection has been demonstrated in vitro, ex vivo, and in vivo
(Konofagou and Hynynen 2003, Maleke et al 2006, Curiel et al 2009a, 2009b, Maleke and Konofagou 2010).
Mechanical properties are usually computed from the measured mechanical response data through model
fitting approaches. The most computationally intensive are full field inverse problem approaches that use
optimization strategies to fit a discretized description of the full field property distributions to all of the available
measured data. Examples include compression elastography, where discretized parameters from the constitutive
equations describing quasi-static deformation of an isotropic linear elastic material (Navier’s equation) with a
plane strain assumption are fitted to the quasi-static strain data (Ophir et al 1991). Nonlinear inversion MR
elastography fits unknown properties from a three-dimensional finite element model of harmonic motion for a
heterogeneous viscoelastic (McGarry et al 2013) or poroelastic (McGarry et al 2015) material to the measured
harmonic motion amplitudes. An inverse problem approach has also been demonstrated for ultrasound pulse
wave imaging, where the unknown compliance distribution and inlet boundary conditions for the 1D pulse
wave equations are fitted to spatiotemporal wall displacement measurements (Mcgarry et al 2016, McGarry et al
2017).
Significantly faster model-fitting approaches have also been demonstrated using simplified models. Direct
inversion MR elastography fits the homogenous form of the harmonic Navier’s equation to motion data, which
has been filtered in an attempt to remove the contribution of the compressional wave. Shear wave elastography

2
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 1. Schematic overview of the HMI setup.

Figure 2. Right: axisymmetric FEM model with infinite elements, with a 100×50 mm section of finite elements, surrounded on
three sides by infinite elements. Left: zoomed-in view of the focus. is the center line of rotational symmetry. The mesh was finely
resolved in the center to accurately capture the wave dynamics near the small focal region, and coarser in the far field regions to reduce
the computational burden where precise modeling is less important. A mesh convergence study confirmed this was an efficient
meshing strategy and the displacement field in the important region surrounding the focus had low discretization error.

3
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

uses the theoretical shear wave propagation speed in either an infinite homogenous isotropic medium
(Gennisson et al 2003) or an infinite homogenous plate surrounded by fluid (Widman et al 2015) to relate the
shear modulus to the observed wave propagation speed. Some methodologies, such as pulse wave imaging
(Meinders et al 2001, Vappou et al 2010), or optical methods (Campo and Dirckx 2011, Campo et al 2016),
amongst others, derive the Young’s modulus of arteries from the velocity of 1D pressure waves propagating
through the structure.
This paper develops, validates, and evaluates an analytical model for the displacement field resulting from
HMI in an infinite homogenous linear elastic material, which allows rapid and accurate computation of the
displacement, strain, and SWV at any point in the material without solving a full-field problem. Additionally, the
effect of violating the model assumptions on the outcome of the analysis is investigated in cases of a free surface
with the HMI focus at varying depth (a half-space, approximating the tissue surface), and for cases where the
HMI focus is inside a stiffer inclusion in a softer background (modeling tumor tissue with a different stiffness
than the surrounding background Konofagou et al 2012). Finally, the FEM model is compared with
experimental HMI results in a gelatin phantom.

2. Methods

2.1. Experimental setup


A schematic illustration of the main components of the HMI system is shown in figure 1. A 2D HMI system was
used in this study, which can be used for tumor detection and HIFU treatment monitoring. For more details,
please consult our group’s previous publications (Hou et al 2014).

2.2. Finite element model


An axisymmetric FEM was generated in ABAQUS (Dassault systèmes, Vélizy-Villacoublay, France), consisting
of a 100 mm diameter, 100 mm high cylinder of finite elements surrounded by infinite elements at the boundary
to absorb reflections. A uniformly distributed 10000 Nm−3 loading was applied across a 1.5×3 mm cylindrical
volume in the center and oscillated sinusoidally, starting from rest at a fixed driving frequency of 50 Hz. The
force voxel, driving frequency and loading are comparable to those used in a clinical setting in, e.g. pancreatic
tissue (Nabavizadeh et al 2020) or breast tissue (Samani et al 2003). Reported stiffness properties of soft tissue
show great variability. We modeled soft tissue as being isotropic elastic with Young’s modulus of 10 kPa, a
Poisson’s ratio of 0.49, and a density of 1000 kg m−3, as reported in e.g. Samani et al (2003) or in Wells and Liang
(2011). The geometry is shown in figure 2. The transient response of the system was computed using the
ABAQUS linear implicit solver. Results were extracted from the ABAQUS output through a Python script
(Python 2.7) and processed using Matlab (Mathworks, Natick, MA, USA).

2.3. Analytical model


The analytical solution for displacement in the axial direction (defined as the z direction) resulting from an
axially directed harmonic point force, Fz , at frequency w, centered at a location (0, 0, 0) in an infinite material is
given by Callé et al (2005) as follows:
⎛⎡ 2 ⎛ z2 ⎞ ⎛ z2 ⎞ ⎤
⎜⎢ z kc2 ⎜3i 2 - i⎟ kc ⎜1 - 3 2 ⎟ ⎥
 Fz ⎜⎢ r 2 ⎝ r ⎠ ⎝ r ⎠ ⎥ ik c r
Uz (x ) = ⎜ + + e
4prw 2 ⎢ r r 2 r 3 ⎥
⎜⎢ ⎥
⎝⎣ ⎦
⎡⎛ z2 ⎞ ⎛ z2 ⎞ ⎛ z2 ⎞⎤ ⎞
⎢ ⎜1 - 2 ⎟ ks2 ⎜i - 3i 2 ⎟ ks ⎜3 2 - 1⎟ ⎥ ⎟
⎢ ⎝ r ⎠ ⎝ r ⎠ ⎝ r ⎠ ⎥ iks r ⎟
+ + + e ⎟. (3)
⎢ r r2 r3 ⎥
⎢ ⎥ ⎟
⎣ ⎦ ⎠

Here, r is the density, and r is the Euclidean distance of the point (x , y , z ) from the point force. ks and kc are the
shear and sound wavenumbers, respectively, given by:

rw 2 rw 2
ks = , kc = . (4)
m M

4
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

For a material with shear modulus, m and p-wave modulus, M. Note that Calle et al has a minor error in the
1
terms with r 2 which did not affect the results of their analysis (Callé et al 2005). However, we have corrected
it here.
Differentiating equation (3) with respect to z gives an analytical expression for the strain in the direction of
du
the force,  zz = dzz , (often referred to as the axial strain in ultrasound):

⎛⎡ 2 ⎛ z2 ⎞ ⎛ z2 ⎞ ⎤ ⎞
⎜⎢ z k 2 ⎜ 3 i - i ⎟ k c ⎜ 1 - 3 ⎟ ⎥ ⎟
⎜⎢ r 2 c ⎝ r2 ⎠ ⎝ r 2 ⎠ ⎥ z ik c r ⎟
⎜⎢ r + + ik e ⎟
r2 r3 ⎥ cr
⎜⎢ ⎥ ⎟
⎜⎣ ⎦ ⎟
⎜ ⎡ ⎤ ⎟
⎜ ⎢ ⎛⎜1 - z ⎞⎟ k 2 ⎛ z ⎞ ⎛ z ⎞
2 2 2
⎜i - 3i 2 ⎟ ks ⎜3 2 - 1⎟ ⎥ ⎟
Fz ⎜+ ⎢ ⎝ ⎟
s
 r2 ⎠ ⎝ r ⎠ ⎝ r ⎠ ⎥ z iks r
 zz (x ) = ⎜ + + iks e ⎟. (5)
4prw 2 ⎜ ⎢ r r 2 r 3 ⎥ r

⎢ ⎥
⎜ ⎣ ⎦ ⎟
⎜ ⎟
⎜ ⎡ 2⎛ z 3z 3 ⎞ ⎛ z z3 ⎞ ⎛ z z 3 ⎞ ⎤ ik c r ⎟
⎜+ ⎢⎣kc ⎜⎝2 r 3 - r 5 ⎟⎠ + 3ikc ⎜⎝2 r 4 - 4 r 6 ⎟⎠ - 3⎜⎝2 r 5 - 5 r 7 ⎟⎠ ⎥⎦ e ⎟
⎜ ⎟
⎜ ⎡ ⎛ z z 3⎞ ⎛ z z 3⎞ ⎛ z z 3 ⎞⎤ ⎟
⎜+ ⎢ - 3ks2 ⎜3 3 - 3 5 ⎟ - iks ⎜8 4 - 12 6 ⎟ + ⎜9 5 - 15 7 ⎟ ⎥ e iks r ⎟
⎝ ⎣ ⎝ r r ⎠ ⎝ r r ⎠ ⎝ r r ⎠⎦ ⎠

A similar expression for the displacement in coordinate direction y, which is perpendicular to the force, is:

⎛⎡ yz 2 yz yz ⎤
k 3i 2 kc -3 2 ⎥
 Fz ⎜⎢ r 2 c
Uy (x ) = ⎜⎢ + r2 + r e ik c r

4prw 2 ⎜⎢ r r r3 ⎥
⎝⎣ ⎦
⎡ yz 2 yz yz ⎤ ⎞
⎢ - r 2 ks - 3i 2 ks
r
3 2⎥
r ⎟
+⎢ + 2
+ 3 ⎥ e s ⎟.
ik r (6)
⎢⎣ r r r ⎥


dUy
And differentiating with respect to y gives the strain along direction y,  yy = dy
, (often referred to as the
lateral strain in ultrasound):

⎛ ⎡ yz 2 yz yz ⎤ ⎞
⎜ ⎢ 2 kc 3i 2 kc -3 2 ⎥ ⎟
y
⎜⎢ r + r2 + r ik e ik c r
⎥ c ⎟
⎜⎢ r r r3 ⎥ r ⎟
⎜⎣ ⎦ ⎟
⎜ ⎡ yz 2 yz yz ⎤ ⎟
⎜ ⎢ - 2 ks - 3i 2 ks 3 2⎥
y ⎟
⎜ + r + r + r ik e iks r ⎟
 Fz ⎢ ⎥ s
 yy (x ) = ⎜ ⎢ r r2 r3 ⎥ r ⎟. (7)
4prw ⎜ ⎣
2 ⎦ ⎟
⎜ ⎡ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎤ ⎟
⎜+ ⎢k 2 ⎜ - 3y z + z ⎟ + 3ikc ⎜ - 4y z + z ⎟ - 3⎜ - 5y z + z ⎟ ⎥ e ik c r ⎟
2 2 2

⎜ ⎣ c ⎝ r5 r3 ⎠ ⎝ r6 r4 ⎠ ⎝ r7 r5 ⎠⎦ ⎟
⎜ ⎟
⎜ ⎡ 2 ⎛ 3y 2z z⎞ ⎛ - 4y 2z z⎞ ⎛ - 5y 2z z ⎞ ⎤ iks r ⎟
⎜+ ⎢ - ks ⎜⎝ - 5 + 3 ⎟⎠ - 3iks ⎜⎝ 6 + 4 ⎟⎠ + 3⎜⎝ 7 + 5 ⎟⎠ ⎥ e ⎟
⎝ ⎣ r r r r r r ⎦ ⎠

A similar expression can be developed for the shear strains  yz or  xz . We develop  yz = 1


2 ( dUz
dy
+
dUy
dz ) but a
similar expression can be found for  xz :

5
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 3. Panel (A): discretization of the total volume in a Cartesian coordinate system. The red cylinder is the force-volume subject to
body forces. The pink square box is a force-voxel inside the force-volume. The material grid-points are blue, and the force grid-points
are red. The distance between the force grid-points is dx, dy, or dz. The query force grid-point inside the force-voxel is indicated with a
green circle. Panel (B): close-up of the force-volume. The force-volume in this study is a 3 mm high cylinder with a diameter of
1.5 mm. The pink square box is a force-voxel inside the force-volume. The query force grid-points inside the force-volume are red.
The query point inside the force-voxel is indicated with a green circle. Force-volume, force-voxel, and total volume are not to scale;
force voxels may be much smaller than the force-volume.

⎛⎡ ⎛ 2 2⎞ ⎛ 2 2⎞ ⎞
⎜ ⎢i kc3 ( z 2) + kc2 ⎜ r - 6 z ⎟ + i kc ⎜ 3 r - 15 z ⎟ ⎟
⎜⎣ ⎝ r ⎠ ⎝ r2 ⎠ ⎟
⎜ ⎟
⎜ - ⎛⎜ 3 r 2 - 15 z 2 ⎞⎟ ⎤ y e ik c r ⎟
⎜ ⎝ ⎥ ⎟
r3 ⎠⎦ r4
⎜ ⎟
 Fz ⎜⎡ ⎛3 2 ⎞ ⎟.
 yz (x ) = r - 6z ⎟
2 (8)
4prw 2 ⎜⎢ 3 ⎛ 1 ⎞ ⎜ ⎛ 3 r 2 - 15 z 2 ⎞ ⎟
⎜⎢i ks ⎜ r 2 - z 2⎟ - ks2 ⎜ 2 ⎟ - i ks ⎜ ⎟⎟
⎜⎢ ⎝2 ⎠ ⎜ r ⎟ ⎝ r2 ⎠⎟
⎜⎣ ⎝ ⎠ ⎟
⎜ ⎛ 2 ⎞ ⎤ ⎟
⎜ + ⎜ 3 r - 15 z ⎟ ⎥ y e iks r
2

⎜ ⎝ ⎠ ⎟
⎝ r 3 ⎦ r 4

The force distribution for a HIFU transducer can be measured by mapping the acoustic pressure field and
applying equation (1) with an estimate of the attenuation of the material. Attenuation estimates can either come
from the literature or use an experimentally determined value (Chen et al 2015). The displacement and strain
fields for a set of displacement query points, xq, for a harmonic ARF force field can be computed using
equations (3)–(8) and the principle of linear superposition. The distributed force field is spatially discretized into
a finite number of volume elements (voxels), and the resulting quantities at the query points for each force voxel
are added together to yield the overall field
Nf

Q tot (x q ) = å Q (x q - x f ) i , (9)
i=1

where Nf is the amount of force voxels, and xf is the centroid of force voxel i. The generalized quantity, Q, could
be Ux, Uy, Uz,  xx ,  yy ,  zz ,  xy ,  xz or  yz .
In this study, the larger surrounding volume is being discretized using a rectangular grid of Ms=Mx*My*Mz
points, or material grid-points, while the force-volume is being discretized using a rectangular grid of
Nf=Nx*Ny*Nz points, or force grid-points. The distance between force grid-points in x-, y- and z-dimension
are dx, dy, and dz, respectively. To calculate the displacement and strain field, each force grid-point is considered
to be the center of a rectangular force-voxel with dimensions dx*dy*dz (see figure 3). Every force-voxel then
generates its own displacement and strain field in the surrounding medium. In order to calculate the
displacement and strain field generated by the entire force-volume, the fields generated by the different force-
voxels are summed according to the superposition principle. The force grid-point is a singularity according to
equations (3)–(8). Due to the finite discretization of the force-volume, this leads to unrealistically high values
when the material grid-point is inside or very close to a force grid-point. Therefore, a 2D median filter in this
study was applied after full assembly of the analytical model.

6
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

2.4. Effect of violation of assumptions


Numerical experiments with FEMs tested the effect of the infinite medium assumption, as well as the
homogeneity assumption on the displacement and strain field.
For the semi-infinite medium test, a FEM mesh with a free surface on the top surface and infinite elements
on the other free sides was created, and the displacement and strain fields calculated for various depths D of the
HMI focus (D=10, 20, 30, 40, 50 mm and ∞). These depths are in line with the measurement depths of in vivo
HMI measurements (Saharkhiz et al 2020). The fields display the maximum value of displacement and strain in
every material grid-point. For each value of D, the CV was calculated as compared to the reference case with
infinite elements on all sides (fully infinite medium) for Uz and  zz . The CVp for a parameter p, at a grid-point mi
is described as:
100 * (preference (m i ) - ptest (m i )) * Ms
CVp(m i ) = , (10)
å i= 1 abs (preference (m i )
Ms

with preference(mi) and ptest(mi) the parameter values at material grid-point mi as calculated in the reference state,
and with violation of assumptions respectively, and Ms the total number of material grid-points in the region of
interest.
For the homogeneity test, a spherical inclusion with a 2× higher stiffness of varying radius R was centered
at the HMI focus (R=0, 3, 5, 7, 10, 15, 20, 30, 40 and ∞ mm). Inclusion sizes in this work are in line with
investigated tumor sizes in in vivo HMI studies (Saharkhiz et al 2020, Hossain et al 2021). For each value of R, CV
was calculated as in equation (10) as compared to the reference case with a homogeneous infinite medium for Uz
and  zz fields to determine the resolution of a single HMI measurement on a small, stiff structure. The fields
display the maximum value of displacement and strain in every material grid-point.
Since an overlap of the force volume between inclusion and surrounding medium possibly complicates
interpretation of the results, the size of the spherical inclusion is limited. Therefore, the minimum radius of the
inclusion (a sphere of 3 mm) is chosen such that the focus (a cylinder of 3×1.5 mm) is completely surrounded
by the inclusion material.

2.5. Validation and convergence analysis of the analytical model


The analytical model was validated with the FEM. In order to quantify the accuracy of the model, the CV was
calculated for the displacement and strain fields as described in equation (11):
The CVp for a parameter p, at a grid-point mi is described as:

100 * (pFEM (m i ) - panalytic (m i )) * Ms


CVp(m i ) = , (11)
å i= 1 abs (pFEM (m i )
Ms

with pFEM(mi) and panalytic(mi) the parameter values at material grid-point mi as calculated by FEM and the
analytical model respectively, and Ms the total number of material grid-points in the region of interest.
The fields display the maximum value of displacement and strain in every material grid-point. This was done
in a sampling grid covering 3 mm in the axial direction and 50 mm in the radial direction, with sampling
resolutions of 0.10 mm. and 0.25 mm, respectively. The radial dimension of the sampling grid was chosen so
that it would match one shear wavelength. Sampling resolution was kept the same in the analytical model and
the FEM, the same cylindrical force distribution as the FEM model was reproduced in the analytical model, and
the mechanical parameters were kept identical. A typical HMI forcing frequency of 50 Hz was applied in both
models. In the analytical model, the resolution of the force discretization was 0.05 mm. In order to calculate the
CV per field, the difference between FEM and the analytical field was divided by the maximum FEM field-value
per material grid-point and expressed as a percentage for every field. The overall accuracy was calculated by
taking the average of the absolute values of the entire CV field. Additionally, apparent SWV fields were compared
between models.
The convergence of the displacement and strain images produced by the analytical model with increasing
force discretization was tested on a sampling grid covering 3 mm in the radial direction and 6 mm in the axial
direction, with 0.1 mm resolution, giving a total of 1830 sampling locations. Sampling resolution was kept the
same in the analytical model and the FEM, the same cylindrical force distribution as the FEM model was
reproduced in the analytical model, and the mechanical parameters were kept identical. A typical HMI forcing
frequency of 50 Hz was applied in both models. In the analytical model, a range of force discretization levels Nf
between 1 and 82 575 was defined and used to discretize a 1.5 mm diameter, 3 mm height cylindrical focus with a
maximum resolution of 0.04 mm. This was done for the entire maximum  xx ,  zz , and  xz field. The fields
display the maximum value of strain in every material grid-point. In order to calculate the CV per field, the
difference between FEM and the analytical field was divided by the maximum FEM field-value per material

7
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 4. Elevational width of the intensity field for two typical imaging transducers (P4-2 and P12-5), compared to the shape of the
displacement field predicted by the analytical model with a typical FUS force distribution.

grid-point and expressed as a percentage for every field. The overall accuracy was calculated by taking the average
of the absolute values of the entire CV field.

2.6. Plane shear wave speed model


The speed of a plane shear wave in a homogeneous, infinite material (see equation (2)) is often used to estimate
the shear modulus from measurements of the local wave propagation speed. The displacements close to a focal
displacement source described by equation (3) do not fit the plane wave model, leading to errors when
equation (2) is applied to estimate m from the apparent SWV or SWVapparent. Equation (3) is written in a form
1
() 1
() 1
()
which separates the near field r 3 , intermediate field r 2 , and far-field r terms of the Greens function, for
both shear waves (e iks r ), and longitudinal waves (e ikc r ) (Callé et al 2005). HMI produces images of the
displacement field over an area typically spanning around one shear wave-wavelength. Therefore, all three fields
are relevant for shear waves. In a nearly incompressible material, ks  kc , therefore only the near field r 3
1
()
longitudinal wave term has an amplitude comparable to the shear waves. The longitudinal wave propagates at a
much faster speed than the shear wave, so its presence causes errors applying equation (2) to the apparent wave
speed. However, the near field component dissipates quickly, so the effect is limited to quite close to the FUS
focal point. Another factor that can cause errors is the phase difference between the intermediate field term and
the near and far-field terms caused by the factor of i in the intermediate field terms. Although all three fields
propagate at the plane wave approximation speed, the three terms dissipate at different rates. Therefore, as r
varies as we move away from the focus, there is a significant phase shift caused by the relative size of the
intermediate field term compared to the near and far-field terms. Both the phase change and near field
longitudinal wave term could cause errors in using the apparent wave speed to estimate the shear modulus using
the planar wave speed in SWV estimation approaches. Using the theory for our analytical model, the
contributions to the displacement field of the shear and longitudinal waves are visualized for the near,
intermediate, and far-field. The fields display the maximum value of displacement in every material grid-point.
Also, an apparent SWV map was made for waves traveling in the x-direction. This map was calculated by
calculating the phase difference dj (in radians) of the transient signal between every subsequent material grid-
point in the x-direction. If the driving frequency ω is known, as well as the difference between the material grid-
points dx, then the SWVapparent in a material grid-point was approximated as:
2p dx w
SWVapparent = . (12)
dj

2.7. Comparison with experimental data


The displacement predictions of simulations using typical FUS force distributions produces spatial
displacement fields with a very narrow peak of maximum temporal displacement, centered around the HMI
focus, whereas experimental HMI images tend to have wider, flatter peaks. There are several reasons for this
discrepancy, including nonlinear stress–strain behavior, viscous losses, blurring of the HMI focus due to wave
speed and attenuation variation throughout the volume, and the size of the beam profile of each element of the
imaging transducer used to take the displacement measurements. HMI measures displacements using 1D cross-
correlation, where the displacement for a given image location is selected as the shift, which maximizes the cross-
correlation of the RF signal between a displaced and reference image. For a measured FUS field, the width of the

8
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 5. Comparison of the  xx ,  zz and  xz for the FEM ((a), (d) and (g) respectively) and the analytical model ((b), (e), and (h),
respectively). In the rightmost panes, the coefficient of variation (CV) comparing FEM and analytical fields is given (panes (c), (f),
and (i)).

displacement caused at the focus is much smaller than the elevational width of typical imaging transducers, as
shown in figure 4.
And it is also smaller than the size of the axial cross-correlation kernel (typically 3 mm), which has an
averaging effect in the axial direction. The function describing the axial cross-correlation for a distribution of
scatterers displaced by some function uz (x ) is given by McAleavey et al (2003):
⎛ 2Fu z (x , y ) ⎞
R (t ) = ò I (x, y )2 cos ⎜⎝Ft + c
⎟ dxdy ,

(13)

where I (x , y ) is the intensity profile of an imaging transducer element, F is the center frequency of the
transducer, c is the speed of sound in the material, and the integral is taken over the lateral (x), and elevational (y)
beam profile of an imaging transducer element. Assuming the profile, I, is much wider in the elevational
direction, a discretized approximation is
N ⎛ 2Fu z ( yk) ⎞
R (t ) = å Ik2 cos ⎜⎝Ftk + c
⎟ Dyk .

(14)
k=1

To simulate cross-correlation based measurement of a given displacement field, compute uz at k points in


the elevational direction. Find the t that maximizes R in equation (13), and then the appropriate displacement
measurement is given by
t c
Ucc = - max . (15)
2
An HMI experiment with a phantom constructed from 50.7 g l−1 of gelatin (50 bloom, MP Biomedicals,
Irvine, CA, USA) with silica powder scatterers was conducted at a frequency of 100 Hz using a P4-2 Imaging
transducer. Typical in vivo HMI is performed at 50 Hz, we used a higher frequency here to observe a full
wavelength across the FOV and compare to the model predictions. Displacements were estimated using the
standard cross-correlation processing pipeline (Chen et al 2015). The pressure field of the FUS transducer was
mapped in water using a fiber optic hydrophone (HFO-690, Onda, Sunnyvale, CA, USA) and used to develop an
axisymmetric description of the FUS force in the focal region, assuming an attenuation of 60 m−1 in equation (1)

9
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 6. Verification of Analytical and FEM models of HMI. The figure displays the time response of the axial strain in the FEM and
analytical model for a driving frequency of 50 Hz, at the locations indicated in the full-field image. The 2D strain map in the illustration
is generated from our analytical model.

Figure 7. (a) Convergence behavior of the analytical model with Ms=1830 displacement sampling points covering a 3×6 mm2
region at a resolution of 1.0 × 10−4 m. A 1.5 mm diameter 3 mm high cylindrical focus was discretized into Nf segments
(Nf=1–82575). Plot a shows the strain difference expressed as CV for òxx, òzz and òxz compared to the FEM for the whole 3×6 mm2
region. (b) Shows the computation time for a serial Matlab implementation running on a 2.4 GHz Intel Xeon E5-2630 CPU per
sampling point.

and correcting for attenuation with depth. This attenuation value is from an unpublished experiment on a
similar phantom using the method described in Chen et al (2015). A shear modulus of μ = 4 kPa was selected for
the analytical model to match the wavelength visible in the experimental images. With a driving frequency F and
a density ρ, the shear modulus μ and the wavelength λ are related through the following equation:

1 m
l= (16)
F r

10 planes of data were generated across the elevational beamwidth, and Ucc was calculated using the elevational
beam profile in figure 4.

10
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 8. An illustration of SWVapparent map in the analytical model (a) and FEM model (b). In (c), a cross-section of (a) along the red
dotted line is shown. As can be observed, the SWVapparent is a function of lateral distance to the oscillating volume. The blue and the
red plot are the SWVapparent as measured in the FEM model, and in analytical, respectively, the green plot indicates the theoretical
plane-wave SWV (SWVpw=1.8 m s−1) as calculated by equation (2).

Table 1. Absolute difference between the FEM and analytical strain ( ) and displacement (U) fields. The average percentage difference
for the whole field and minimum (Min) and maximum (Max) percentage difference of all the material grid-points are given, both for
the 41 940 and 82 575 sampling points of the focus. Additionally, the time per displacement evaluation is given.

41 940 82 575
Number of force voxels
0.22 0.39
time per displacement evaluation (s)
Average (%) Min (%) Max (%) Average (%) Min (%) Max (%)

Ux 1.0 4.4E-05 2.6 0.24 0.00015 0.58


Uz 1.2 0.0017 2.4 1.0 0.22 1.2
 xx 0.51 0.00044 4.4 0.14 4.2E-05 6.8
 zz 0.45 1.8E-05 3.2 0.12 4.3E-05 6.3
 xz 0.50 0.00031 4.7 0.15 1.3E-05 1.5

3. Results

3.1. Validation and convergence analysis of the analytical model


The accuracy of the analytical model was checked by computing the CV of the displacement and strain as
compared with the FEM (see figures 5 and 6). The average CV between both models is less than 0.51% for the
strains and less than 1.2% for the displacements (see table 1). Maximum displacements are between 1 and 2 μm,
which is in agreement with displacements reported in vivo, in e.g. Saharkhiz et al (2020).
A convergence study was performed with the increasing of the resolution for the force-volume in the
analytical model, as compared to the FEM model. When the force-volume is sampled at a resolution as high as
4 × 10−5 m, the CV can be decreased below 0.15% and 1.1% for  zz and Uz respectively for a computation time
of 0.39 s per displacement value (see figure 7(a), see table 1) or a computational cost increase of 76%.
The slope of the plot indicates that computation time for the analytical model is proportional to O (Nd )3 , as
would be expected (see figure 7(b)) as every displacement computation is independent of the others. The
computation time is also linearly related to the number of points in the strain image, as each point is
independent of the others so that a single value can be computed very quickly, even with high-resolution force
discretization.
The SWVapparent field was also compared between models, as well as with the theoretical plane-wave SWVpw
(1.8 m s−1 as defined by equation (2)) (see figure 8). As observed, the SWVapparent is a function of lateral distance
to the oscillating volume for both the FEM and the analytical model. The SWVapparent field, as computed from
FEM and the analytical model, differs by up to 0.57 m s−1, with a mean absolute difference of 0.11±0.07 m s−1.
The FEM and analytical SWVapparent follow the same general trends. The oscillation of the FEM model is likely
from reflections due to the imperfect absorption of incoming waves by the infinite elements at the boundary (see
figure 8). At focus depth, the maximum observed difference between SWVpw and SWVapparent is 1.2 m s−1 for
the FEM and 0.99 m s−1 for the analytical model. These differences are due to the influence of the near field
longitudinal wave and different dissipation rates of the near, intermediate, and far-field terms. The SWV
simulated in our model, corresponds to the SWV observed in in vivo measurements, as SWV reported in e.g.
(Toprak et al 2019) varies from 1.59 m s−1 in healthy soft tissue, to 6.51 m s−1 in cancerous tissue.

11
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 9. Effect of nearby free surface on the FEM model. Panel (a) illustrates the axisymmetric geometry, with one free surface and
infinite elements on the other sides. The distance D, i.e. the distance from the HMI focus to the surface, is varied. Panel (b) plots the
coefficient of variation (CV) compared to the reference case with infinite elements on all sides (fully infinite medium). Panel (c) shows
the maximum displacement for each value of D compared the reference infinite medium case (D=∞). Panel (d) is the axial
displacement field in the reference infinite medium case, and (e) and (f) show the displacements for D=10 mm and D=50 mm.
Panels (g)–(k) show the difference Φ between each semi-infinite model and the reference infinite medium.

Figure 10. Effect of spherical inclusion centered at the focus on maximum displacement and strain. The geometry is displayed in (a)
with the inclusion with radius R in an infinite medium. (b) displays the axial displacements with the different radii included in this
analysis (R=0–∞). (c) and (d) show the evolution with increasing R of the maximum displacement and strain, respectively.

12
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 11. Illustration of the contribution of the different axial displacement fields of the near, intermediate, and far shear terms ((a),
(c), and (e), respectively) and longitudinal terms ((b), (d), and (f), respectively) as derived from the analytical model in a region up to
50 mm laterally from the oscillating volume. Displacements are displayed as the logarithm of the absolute displacement value for each
field. Note that the axes steps are not equal in order to show the far and near fields. The wavefronts are spherical on equal axes.

3.2. Effect of violation of assumptions


Figure 9 shows the displacement difference Φ between an infinite medium model and a model with a free surface
at the top for various depths of HMI focus. HMI displacement fields are affected by proximity to a free surface,
and the infinite medium approximation becomes accurate to within 4% at a focal depth of 50 mm. The variation
in maximum displacement amplitude is small (on the order of 10%); therefore, using the maximum
displacement as a surrogate for relative stiffness is not strongly affected by focal depth.
For cases when the HMI focus is inside a spherical inclusion shown in figure 10, the maximum displacement
approaches the displacement of an infinite material at the inclusion stiffness to within 10% for inclusions larger
than 10 mm. The maximum strain reacts faster, reaching the infinite inclusion value for inclusions as small
as 3 mm.

13
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

Figure 12. Left: FUS force distribution computed from a mapped intensity field of the FUS transducer used for the experiment. This
field was rotated to form an axisymmetric force distribution for the analytical model. Right: comparison of the real and imaginary
components of the axial displacement Uz and strain  zz for the experiment and predictions of the analytical model.

3.3. Analytical model terms


As can be seen in figure 11, for a nearly incompressible material (kc  ks ), the intermediate and far-fields of
longitudinal waves are negligible (orders of magnitude smaller than other fields at all locations). However, the
near field term in equation (3) is not multiplied by kc , so it is of comparable size to the shear terms near the focus.

3.4. Comparison with experimental data


The left panel of figure 12 shows the force distribution generated from the pressure mapping of a HIFU
transducer, which was converted into an axisymmetric 3D force profile for the simulation. The real and
imaginary parts of the experimental and analytical model incremental displacement and strain are compared
(experimental displacements were Fourier transformed to recover the 100 Hz complex amplitude to compare
with the model). The images are quantitatively similar, and the model reproduces the major features of the
experimental images. There is some discrepancy in the absolute values, likely due to an incorrect assumption of
attenuation in the material, and the model predicts a relatively higher displacement right at the focus compared
to the propagating wavefronts, which is likely due to unmodelled effects such as focal blurring, and nonlinear
stress–strain behavior.

4. Discussion and conclusion

HMI has demonstrated its use for elastography through direct monitoring of the displacement field, the strain
field, or the SWV (Konofagou and Hynynen 2003, Maleke et al 2006, Curiel et al 2009a, 2009b, Maleke and
Konofagou 2010). It is also used to monitor the progress of ablation procedures, such as is the case in HIFU
(Han et al 2016). However, for in vivo applications, deviations from idealized cases such reflections from
inhomogeneities and free surfaces make an accurate quantitative estimation of tissue parameters difficult.
Therefore, in general, it is necessary to use a model-based strategy to infer the material property distribution
from the measured tissue response (Glaser et al 2012).
In this work, both a FEM and an analytical model are presented and compared. FEM’s are used extensively
for modeling inhomogeneous tissues with arbitrary geometry. However, the model-building and computation
time associated with this technique can be costly, especially in transient problems. Model-based inversions to
recover the properties through an optimization approach increase the cost even further because multiple
solutions of the FEM model are required as the model parameters are updated.

14
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

The FEM introduced in this work uses an axially symmetric geometry with infinite elements at the
boundaries. Modifications are easily added to the geometry for quantitative and qualitative investigations. For
example, we demonstrated in figure 9 that the distance to the surface has minimal impact on the estimated
deformations and strains compared to the infinite medium approximation for focal depths greater than 50 mm.
The mismatch of the free surface model with the infinite approximation is not linearly correlated to the focus
depth, likely due to resonance (the wavelength of shear waves is ∼= 37 mm). When the focus is close to a free
surface, variations of up to 25% can occur, which demonstrates that the infinite medium model has to be applied
carefully to in vivo situations where the focus can be close to an interface such as the skin surface, the ventral
cavity or bone. In a non-homogeneous situation, where the focus is inside a stiff spherical inhomogeneity with a
radius R (R=0–∞ mm) and Young’s modulus contrast of 2× in a softer surrounding tissue, strain
measurements converge to values associated with the stiff inclusion as its volume increases. The measured
strains and displacements indicate the local stiffness at the focus with a resolution on the order of 6 mm, which is
within the range of 1.2–9 mm that has been reported with HMI in ex vivo and in vivo breast tumor detection
(Saharkhiz et al 2020, Hossain et al 2021).
One major disadvantage of FEM models is that the full-field displacements must be computed even if only a
few displacements are required. The analytical model presented in this study can compute displacements and
strains at any location in the model from an arbitrary harmonic HIFU force field. The model was validated
against the FEM and showed a maximum difference of less than 5% for the strains and less than 3% for the
displacements (see figure 5). In the time domain, there is a good agreement between models (see figure 6).
()1 1
()
Equation (3) is written in a form which separates the near field r , intermediate field r 2 , and far-field r 3
1
()
terms of the Greens function, for both shear waves (e iks r ), and longitudinal waves (e ikc r ) (Callé et al 2005). Images
of each of the six fields (near, intermediate, and far-fields for both shear and longitudinal waves) are presented in
figure 11 for a typical HMI frequency of 50 Hz. All of the individual fields are spherical; however, the overall field
appears ellipsoidal in the near-intermediate region. This is because the intermediate field term is primarily in the
axial direction, and it is 90 degrees out of phase with the other fields due to the factor of i in the equation.
Each field is dissipating at a different rate (determined by the exponent of r ). Therefore, the summed field is
changing phase in the axial direction as it propagates away from the origin. Also, for a nearly incompressible
material (kc  ks ), the intermediate and far-fields of longitudinal waves are negligible. However, the near field
term is not multiplied by kc , so it is of comparable size to the shear terms near the focus (see figure 11(a)), but it
propagates at a much faster velocity. Both the phase change and near field longitudinal wave term can cause
errors in SWV estimation approaches. A demonstration of this effect can be observed in figure 8, as the
SWVapparent field evolves as a function of the lateral distance to the HMI focus, with more substantial deviations
from the plane wave speed (see equation (2)) as we move through the intermediate and near field toward the
focus.
In a convergence analysis, comparing the analytical model with the FEM, the difference between models can
be decreased below 1.1% and 0.15% for the displacements and strains, respectively, for a computation time of
0.39 s per displacement evaluation (see figure 7).
The low computational cost of the analytical model is a strong advantage when using an optimization
routine: the analytical model allows rapid estimation of the material properties as based on the observed
parameters such as the SWV field, displacement field, and strain field. Moreover, the analytical model allows the
computation of single field values without having to build an entire model, potentially reducing computation
cost. The model was able to reproduce the major features of an HMI experiment in a phantom, as illustrated in
figure 12. The most challenging region to model is right at the focus due to very strong displacement and strain
gradients, which are averaged by the ultrasound imaging procedure, and must be included in the model.
Considering the elevational profile of the imaging transducer helped, but other effects such as focal blurring,
viscous behavior, and stress–strain nonlinearity are more challenging to model. The SWV analysis in figure 8
also demonstrated larger deviations from the idealized plane wave speed close to the focus. These findings
suggest that more accurate quantitative stiffness values may result from analyzing regions further from the focus,
rather than directly.
Close to the HMI focus, the measured SWV will deviate from the SWV as predicted by the plane wave
equation, even in idealized conditions such as those presented in this work (see figure 8). The general limitations
of SWV estimation and the effects of non-ideal measurement situations inherent to a real-life setup are well
documented. If the simplifying assumptions used to derive plane wave SWV apply to the measurement
situation, SWV maps would accurately model underlying tissue stiffness. However, this is not always the case,
e.g. due to radially propagating waves from a point-like source, tissue nonlinearity, transducer compression, or
tissue heterogeneities (Shiina et al 2015).
The analytical model represents an ideal situation of well-delineated HMI focus in an infinite homogeneous
medium. As can be observed in our experiments, deviations from the idealized assumptions of a homogeneous

15
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

infinite medium can cause errors in the displacement and strain fields (see figures 9 and 10). In future studies, we
intend to evaluate and validate more complex analytical models, rendering more realistic conditions, including a
free moving surface (a half-space Banerjee and Mamoon 1990) and a heterogeneous material structure (e.g. a
poroelastic half-space Zheng et al 2013 or a layered poroelastic half-space Zheng et al 2013). We also intend to
compare model-based mechanical property estimates from experimental data using the analytic model against
competing approaches.

Acknowledgments

This research was supported by the Research Foundation Flanders (FWO) (grant number 12F2317N).

ORCID iDs

Matthew D J McGarry https://orcid.org/0000-0001-8430-800X


Adriaan Campo https://orcid.org/0000-0003-0219-6054

References
Alizad A, Whaley D H, Urban M W, Carter R E, Kinnick R R, Greenleaf J F and Fatemi M 2012 Breast vibro-acoustography: initial results
show promise Breast Cancer Res. 14 R128
Banerjee P K and Mamoon S M 1990 A fundamental solution due to a periodic point force in the interior of an elastic half-space Earthq. Eng.
Struct. Dyn. 19 91–105
Bercoff J, Tanter M and Fink M 2004 Supersonic shear imaging: a new technique for soft tissue elasticity mapping IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 51 396–409
Callé S, Remenieras J-P, Bou Matar O, Hachemi M E and Patat F 2005 Temporal analysis of tissue displacement induced by a transient
ultrasound radiation force J. Acoust. Soc. Am. 118 2829–40
Campo A and Dirckx J 2011 Dual-beam laser Doppler vibrometer for measurement of pulse wave velocity in elastic vessels Proc. SPIE 8011
80118Y
Campo A, Heuten H, Goovaerts I, Ennekens G, Vrints C and Dirckx J 2016 A non-contact approach for PWV detection: application in a
clinical setting Physiol. Meas. 37 990–1003
Catheline S, Souchon R, Rupin M, Brum J, Dinh A H and Chapelon J-Y 2013 Tomography from diffuse waves: passive shear wave imaging
using low frame rate scanners Appl. Phys. Lett. 103 014101
Chen H et al 2015 Harmonic motion imaging for abdominal tumor detection and high-intensity focused ultrasound ablation monitoring: an
in vivo feasibility study in a transgenic mouse model of pancreatic cancer IEEE Trans. Ultrason. Ferroelectr. Freq. Control 62 1662–73
Chen J, Hou G Y, Marquet F, Han Y, Camarena F and Konofagou E 2015 Radiation-force-based estimation of acoustic attenuation using
harmonic motion imaging (HMI) in phantoms and in vitro livers before and after HIFU ablation Phys. Med. Biol. 60 7499–512
Curiel L, Chopra R and Hynynen K 2009a In vivo monitoring of focused ultrasound surgery using local harmonic motion Ultrasound Med.
Biol. 35 65–78
Curiel L, Huang Y, Vykhodtseva N and Hynynen K 2009b Focused ultrasound treatment of VX2 tumors controlled by local harmonic
motion Phys. Med. Biol. 54 3405–19
Deffieux T, Montaldo G, Tanter M and Fink M 2009 Shear wave spectroscopy for in vivo quantification of human soft tissues visco-elasticity
IEEE Trans. Med. Imaging 28 313–22
Fatemi M and Greenleaf J F 1999 Vibro-acoustography: an imaging modality based on ultrasound-stimulated acoustic emission Proc. Natl
Acad. Sci. 96 6603–8
Fatemi M, Wold L E, Alizad A and Greenleaf J F 2002 Vibro-acoustic tissue mammography IEEE Trans. Med. Imaging 21 1–8
Gennisson J-L, Catheline S, Chaffaı̈ S and Fink M 2003 Transient elastography in anisotropic medium: application to the measurement of
slow and fast shear wave speeds in muscles J. Acoust. Soc. Am. 114 536–41
Glaser K J, Manduca A and Ehman R L 2012 Review of MR elastography applications and recent developments J. Magn. Reson. Imaging 36
757–74
Han Y, Wang S, Hibshoosh H, Taback B and Konofagou E 2016 Tumor characterization and treatment monitoring of postsurgical human
breast specimens using harmonic motion imaging (HMI) Breast Cancer Res. 18 46
Hossain M, Saharkhiz N and Konofagou E 2021 Feasibility of harmonic motion imaging using a single transducer: in vivo imaging of breast
cancer in the mouse model and human subjects IEEE Trans. Med. Imaging (https://doi.org/10.1109/TMI.2021.3055779)
Hou G Y et al 2014 Sparse matrix beamforming and image reconstruction for 2D HIFU monitoring using harmonic motion imaging for
focused ultrasound (HMIFU) with in vitro validation IEEE Trans. Med. Imaging 33 2107–17
Konofagou E E, D’hooge J and Ophir J 2002 Myocardial elastograph—a feasibility study in vivo Ultrasound Med. Biol. 28 475–82
Konofagou E E and Hynynen K 2003 Localized harmonic motion imaging: theory, simulations and experiments Ultrasound Med. Biol. 29
1405–13
Konofagou E E, Maleke C and Vappou J 2012 Harmonic motion imaging (HMI) for tumor imaging and treatment monitoring Curr. Med.
Imaging Rev. 8 16–26
Lizzi F L et al 2003 Radiation-force technique to monitor lesions during ultrasonic therapy Ultrasound Med. Biol. 29 1593–605
Maleke C and Konofagou E E 2010 In vivo feasibility of real-time monitoring of focused ultrasound surgery (FUS) using harmonic motion
imaging (HMI) IEEE Trans. Biomed. Eng. 57 7–11
Maleke C, Pernot M and Konofagou E E 2006 Single-element focused ultrasound transducer method for harmonic motion imaging
Ultrason. Imaging 28 144–58
McAleavey S, Menon M and Elegbe E 2009 Shear modulus imaging with spatially-modulated ultrasound radiation force Ultrason. Imaging
31 217–34

16
Phys. Med. Biol. 66 (2021) 075017 M D J McGarry et al

McAleavey S A, Nightingale K R and Trahey G E 2003 Estimates of echo correlation and measurement bias in acoustic radiation force
impulse imaging IEEE Trans. Ultrason. Ferroelectr. Freq. Control 50 631–41
McGarry M et al 2013 Including spatial information in nonlinear inversion MR elastography using soft prior regularization IEEE Trans. Med.
Imaging 32 1901–9
Mcgarry M, Li R, Apostolakis I, Nauleau P and Konofagou E E 2016 An inverse approach to determining spatially varying arterial
compliance using ultrasound imaging Phys. Med. Biol. 61 5486–507
McGarry M, Nauleau P, Apostolakis I and Konofagou E 2017 In vivo repeatability of the pulse wave inverse problem in human carotid
arteries J. Biomech. 64 136–44
McGarry M D J et al 2015 Suitability of poroelastic and viscoelastic mechanical models for high and low frequency MR elastography Med.
Phys. 42 947–57
Meinders J M, Kornet L, Brands P J and Hoeks A P G 2001 Assessment of local pulse wave velocity in arteries using 2D distension waveforms
Ultrason. Imaging 23 199–215
Muthupillai R, Lomas D J, Rossman P J, Greenleaf J F, Manduca A and Ehman R L 1995 Magnetic resonance elastography by direct
visualization of propagating acoustic strain waves Science 269 1854–7
Nabavizadeh A et al 2020 Noninvasive Young’s modulus visualization of fibrosis progression and delineation of pancreatic ductal
adenocarcinoma (PDAC) tumors using harmonic motion elastography (HME) in vivo Theranostics 10 4614–26
Nightingale K, Soo M S, Nightingale R and Trahey G 2002 Acoustic radiation force impulse imaging: in vivo demonstration of clinical
feasibility Ultrasound Med. Biol. 28 227–35
Ophir J, Céspedes I, Ponnekanti H, Yazdi Y and Li X 1991 Elastography: a quantitative method for imaging the elasticity of biological tissues
Ultrason. Imaging 13 111–34
Redmond D P and Hegge F W 1985 Observations on the design and specification of a wrist-worn human activity monitoring system Behav.
Res. Methods, Instrum., Comput. 17 659–69
Saharkhiz N et al 2020 Harmonic motion imaging of human breast masses: an in vivo clinical feasibility Sci. Rep. 10 1–13
Samani A, Bishop J, Luginbuhl C and Plewes D B 2003 Measuring the elastic modulus of ex vivo small tissue samples Phys. Med. Biol. 48
2183–98
Sandrin L et al 2003 Transient elastography: a new noninvasive method for assessment of hepatic fibrosis Ultrasound Med. Biol. 29 1705–13
Sarvazyan A P, Rudenko O V, Swanson S D, Fowlkes J B and Emelianov S Y 1998 Shear wave elasticity imaging: a new ultrasonic technology
of medical diagnostics Ultrasound Med. Biol. 24 1419–35
Shigao C et al 2009 Shearwave dispersion ultrasound vibrometry (SDUV) for measuring tissue elasticity and viscosity IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 56 55–62
Shiina T et al 2015 WFUMB guidelines and recommendations for clinical use of ultrasound elastography: I. Basic principles and terminology
Ultrasound Med. Biol. 41 1126–47
Song P, Zhao H, Manduca A, Urban M W, Greenleaf J F and Chen S 2012 Comb-pBreast vibro-acoustographyush ultrasound shear
elastography (CUSE): a novel method for two-dimensional shear elasticity imaging of soft tissues IEEE Trans. Med. Imaging 31
1821–32
Suomi V, Han Y, Konofagou E and Cleveland R O 2016 The effect of temperature dependent tissue parameters on acoustic radiation force
induced displacements Phys. Med. Biol. 61 7427–47
Toprak N, Yokus A, Gündüz M and Akdenız H 2019 Histopathology and elastography discordance in evaluation of breast lesions with
acoustic radiation force impulse elastography Pol. J. Radiol. 84 224–33
Torr G R 1984 The acoustic radiation force Am. J. Phys. 52 402–8
Vappou J, Luo J and Konofagou E E 2010 Regional measurement of arterial stiffness using pulse wave imaging (PWI): phantom validation
and preliminary clinical results 2010 IEEE Int. Ultrasonics Symp. (San Diego, CA, 11-14 Oct. 2010) pp 1332–5
Weaver J B et al 2012 Brain mechanical property measurement using MRE with intrinsic activation Phys. Med. Biol. 57 7275–87
Wells P N T and Liang H D 2011 Medical ultrasound: imaging of soft tissue strain and elasticity J. R. Soc. Interface 8 1521–49
Widman E et al 2015 Shear wave elastography plaque characterization with mechanical testing validation: a phantom study Phys. Med. Biol.
60 3151–74
Zheng P, Ding B, Zhao S-X and Ding D 2013 3D dynamic Green’s functions in a multilayered poroelastic half-space Appl. Math. Modelling
37 10203–19
Zheng P, Zhao S-X and Ding D 2013 Dynamic Green’s functions for a poroelastic half-space Acta Mech. 224 17–39

17

You might also like