IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO.

8, AUGUST 2015

1969

Denoising of Contrast-Enhanced Ultrasound Cine

Sequences Based on a Multiplicative Model
Avinoam David Bar-Zion , Charles Tremblay-Darveau, Melissa Yin, Dan Adam, Member, IEEE,
and F. Stuart Foster, Senior Member, IEEE

AbstractBackground: Speckle noise is an inherent characteristic of dynamic contrast-enhanced ultrasound (DCEUS) movies and
ultrasound images in general. Speckle noise considerably reduces
the quality of these images and limits their clinical use. Currently,
temporal compounding and maximum intensity persistence (MIP)
are among the most widely accepted processing methods enabling
the visualization of vasculature using DCEUS. Goal: A different
approach has been used in this study, in order to improve the noise
removal, while enabling the investigation of CEUS dynamics. Methods: A multiplicative model for the formation of DCEUS speckled
images is adopted and the log-transformed cines are processed. A
preprocessing step was performed, locally removing low value outliers. Due to the fast-changing spatial distribution of microbubbles
inside the vasculature, the noise in consecutive DCEUS frames is
independent, facilitating its removal by temporal denoising. Noise
reduction is efficiently achieved by wavelet denoising, in which the
signals wavelet coefficients are thresholded and small-value noiserelated coefficients are discarded. The main advantage of using
wavelet denoising in the present context is its ability to estimate
ultrasound contrast agents (UCA) concentration over time adaptively, without assuming a model or predefining the signals degree
of smoothness. The performance of wavelet denoising was compared against MIP, temporal compounding, and Log-normal model
fitting. Results: Phantom experiments showed improved SNR, using wavelet denoising over a wide range of UCA concentrations
(MicroMarker, 0.0011%). In the in vivo tests, improved noise removal was achieved, reflected by a significantly lower coefficient of
variation in homogeneous vascular regions (p < 0.01).
Index TermsContrast-enhanced ultrasound, denoising, multiplicative noise, outlier-resistant estimation.

I. INTRODUCTION
T is well known that bloodtissue contrast in ultrasound
scans can be improved by injecting the patient with ultrasound contrast agents (UCA) and using contrast-specific pulse
sequences [1]. The increased echogenicity and typical radii

Manuscript received October 13, 2014; revised January 13, 2015; accepted
February 14, 2015. Date of publication February 27, 2015; date of current
version July 15, 2015. This work was supported by the Canadian Institutes of
Health Research, the Terry Fox Foundation, the Ontario Research Fund, and
VisualSonics Inc. Asterisk indicates corresponding author.
A. D. Bar-Zion is with Department of Biomedical Engineering,
TechnionIsrael Institute of Technology, Haifa 32000, Israel (e-mail:
barz@tx.technion.ac.il).
C. Tremblay-Darveau is with the Department of Medical Biophysics, University of Toronto.
M. Yin is with the Sunnybrook Health Sciences Center.
D. Adam is with the TechnionIsrael Institute of Technology.
F. S. Foster is with the Sunnybrook Health Sciences Center, and also with the
University of Toronto.
This paper has supplementary downloadable material available at http://
ieeexplore.ieee.org (File size: 7.17 MB).
Digital Object Identifier 10.1109/TBME.2015.2407835

range makes UCAs ideal for the imaging of blood flow. When
imaged at low intensities (mechanical index < 0.05), microbubbles oscillate stably and circulate in the vasculature for long
periods of time, typically many minutes, before collapsing or
being absorbed via endocytosis. The extended circulation of
contrast agents inside the vasculature increases the importance
of the wash-out phase in scans incorporating a bolus injection
of UCA, and enables the use of targeted microbubbles that
bind to specific antibodies [2]. The use of DCEUS imaging
for the quantification of blood perfusion relies on the linear
correlation between UCA concentration and the received harmonic ultrasound intensity. As presented in [3], with a sufficient dynamic range, this linearity holds over three orders of
dilution.
The ability to detect echoes of nonlinear bubble oscillations
and to suppress linear signals from tissue is essential for perfusion imaging. Several multipulse sequences were proposed for
that purpose [4]. In the pulse inversion technique, the polarity of
every two consecutive pulses is altered, preserving the even harmonics. Alternatively, amplitude modulation can be used for the
separation of odd harmonics. A method that manipulates both
amplitudes and phases has been shown to improve UCA detection sensitivity [4]. Using these multipulse sequences, UCA can
be imaged with sufficient contrast to tissue ratio.
Speckle noise is an inherent feature of DCEUS and of ultrasound imaging in general, as it is a coherent imaging modality.
This noise results from the presence of many small scatterers
within each resolution cell, defined by the 6 dB of the main
beam profile in the axial, lateral, and out of plane orientations.
The combination of these reflections creates complicated constructive and destructive interference patterns in the echo signal,
named speckle noise. Speckle noise further reduces the contrast
and resolution of DCEUS and B-mode images and limits their
diagnostic value.
Therefore, speckle reduction has been the focus of many studies aiming to improve the quality of B-mode ultrasound imaging. Following the multiplicative model of speckle formation
first presented in [5], the log transform can be used to convert
the multiplicative noise into additive noise. This additive noise
component can be removed using a variety of classical denoising
methods, for example, wavelet denoising [6], [7]. The underlying assumption of this family of methods is that the resulting
log-compressed images contain an independent identically distributed (i.i.d.) Gaussian noise. In [8], the authors emphasized
both theoretically and experimentally that these assumptions are
generally wrong for log-transformed B-mode images. They presented a preprocessing step that reduced the spatial correlation
of the noise and removed noise outliers. This step produced noise

0018-9294 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

1970

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO. 8, AUGUST 2015

that better resembled an i.i.d. Gaussian noise and significantly

improved the performance of the speckle removal process.
Similar to the case of B-mode images, contrast-enhanced ultrasound scans suffer from speckle noise. There are two main
differences between the speckle patterns that characterize these
two imaging modes. First, due to the resonance phenomena,
the signal received from UCA originates mainly from a subpopulation of microbubbles with radii matching the excitation
frequency. The statistics of the signal received from these bubbles depends on the number of relevant bubbles within each
resolution cell, at a given time. Second, the speckle pattern
in DCEUS images is influenced by the position of the contrast
agents, and their movement with the blood flow over time, unlike
speckles in B-mode images. Moreover, considering typical values of mean flow velocity in arterioles (5 mm/sec), frame rates
(20 frames/s), and the dimension of the resolution cell (300 m
in the axial direction at 18 MHz, 3 cycles), one may assume the
spatial distribution of contrast agents in consecutive frames to
be independent. Therefore, instead of attempting to remove the
structured noise in the spatial domain, the aim here is denoising
the signal in the temporal domaina much easier task.
The work of Sureshkumar [9] was among the first studies
to report on the statistics of CEUS envelope and intensity signals. It was found that for high enough UCA concentrations,
the signals follow Rayleigh statistics. For lower concentrations, the variability of the measurements increased, indicating
deviation from the Rayleigh model. These findings are in
agreement with the results presented in [10] and [11]. The
K-distribution was proposed as a model that can describe the signal and its variability over a wide range of concentrations. This
model was not yet validated. Moreover, since this research was
performed with first harmonic signals, additional study is needed
to examine the validity of the K-distribution model for harmonic
CEUS measurements acquired using multipulse sequences.
Maximum intensity persistence (MIP) is a widely accepted
processing method that enables the visualization of the vasculature from noisy DCEUS ultrasound scans. MIP operates by
calculating the maximum of the intensity signal over time, for
each pixel. By doing so, the vasculature morphology is revealed
[12]. MIP has been successfully used in the delineation of blood
vessels, facilitating the classification of prostate tumors [13],
focal hepatic lesions [12], and breast tumors [14]. In addition,
few papers tried to evaluate perfusion parameters from MIP
movies (for example, peak enhancement in [15]). Nevertheless,
MIP suffers from a few well-known limitations. First, the accumulated maximum intensity over time includes a considerable
structured noise and artifacts. In addition, since MIP is exclusively based on the highest intensity values, it is very sensitive
to motion and every small movement smears the image. Finally,
the wash-out phase, which is characterized by decreasing intensity values, cannot be displayed in MIP cines. As a result, MIP
is still used as a qualitative method and no MIP-based automatic
tool has found its way to the clinic.
Another possible approach to the removal of the speckle noise
from DCEUS scans is temporal compounding. In this context,
temporal compounding is defined as the application of a linear
filter (e.g., FIR filter or an averaging window) to the DCEUS
data in the time domain. For example, averaging of several 3-D

CEUS scans taken during the wash-in phase of a bolus injection,

[16] can be considered a temporal compounding procedure. An
obvious limitation of this approach is that a priori assumption
of the signals degree of smoothness is needed in order to set
the cut-off frequency. Cut-off frequency selection is not trivial
since the signal and the noise can have overlapping supports in
the frequency domain.
Considering the mentioned limitations of current methods,
the aim of this study was to improve the noise removal and enhance the appearance of vascular structures, while enabling the
investigation of local CEUS dynamics. The proposed approach
adopts a multiplicative model for speckle formation in DCEUS
images. It is then shown that the enhancement of vascular structures can be redefined as a denoising problem. The structure
of log-transformed DCEUS signals and the statistics of their
inherent speckle noise are studied to determine an appropriate
denoising method. Since the speckle noise in DCEUS images
is characterized by a heavy tailed distribution, a preprocessing
procedure is proposed for removal of the noise outliers, resulting in a semi-Gaussian noise. A variety of classic denoising
algorithms, optimized for Gaussian noise, can then be used.
This study focuses on the use of wavelet denoising for the
speckle noise rejection in DCEUS cines. The general procedure
for an additive noise removal, by applying a nonlinear threshold to the coefficients of the wavelet transform of the measured
signals, has been developed by Donoho and Johnstone [17],
[18]. The application of wavelet denoising to signals contaminated by the non-Gaussian noise was explored in [19]. The
main advantage of using wavelet shrinkage in our context is its
ability to estimate signals adaptively without assuming a model
or predefining the signals degree of smoothness.
II. MULTIPLICATIVE NOISE MODEL FOR CEUS MOVIES
The main objective of this study is to improve upon current methods of DCEUS signal processing by representing this
problem in the framework of signal denoising. To do so, a model
describing DCEUS signals in both the spatial domain and the
time domain and an understanding of the statistics of the noise
are required.
A. Statistical Models of Speckle Noise in Contrast-Enhance
Ultrasound Scans
Considering the polydispersed physical characteristics of
commercial UCA microbubbles and the variable number of
microbubbles imaged, it is useful to adopt a statistical model
for DCEUS signals. First-order statistics of B-mode images
envelope signal depends on the acoustic homogeneity of the tissue. Similarly, the statistics of CEUS signals primarily depends
on the concentration of relevant microbubbles subpopulations
within the resolution cell and the relative scattering from each of
these bubbles. Based on the analogy between speckle formation
in CEUS and B-mode imaging, models for speckle statistics in
B-mode images are examined in this study, and their validity for
CEUS signals is assessed.
In order to understand the statistics of dephased echoes, the
spatial distribution of the microbubbles should be evaluated.
Commonly, a uniform spatial distribution of microbubbles is

BAR-ZION et al.: DENOISING OF CONTRAST-ENHANCED ULTRASOUND CINE SEQUENCES BASED ON A MULTIPLICATIVE MODEL

assumed [9]. For cases of high concentrations of scatterers

distributed uniformly, the envelope signal is known to follow
the Rayleigh distribution [20]. The speckle amplitude SNR is
an indicator of the speckle contrast, defined as
m1

(1)
r= =

m2 m21
where m is the moment of the corresponding distribution. Inserting the moments of the Rayleigh distribution, we
get r = 1.91 [20]. This constant ratio between the signal (expected value) and the noise (standard deviation) indicates that
the speckle noise is multiplicative for high UCA concentrations.
For the cases where low doses of UCA are injected, or poorly
perfused tissues are scanned, the microbubble concentrations
can be quite low. Many different statistical distributions were
used to describe cases in which the concentration of scatterers
was low, or they could not be assumed to be independently
distributed. The K-distribution will be used in this study, as
proposed in [9], for the modeling of CEUS signals over a large
range of concentrations
 
bA
2b
K 1 (bA) , A 0
(2)
p( A| , b) =
() 2
where is the shape parameter of the K-distribution, E{A2 } is
the second moment of the amplitude, is the gamma function,
and the function Kx () is the modified Bessel
 function of the
second kind with order x. Parameter b = 2 /E{A2 } serves
as a scale parameter. In [21], parameter was shown to be
associated with the concentration of effective reflectors
= ( + 1) Ns ,

> 1

(3)

where Ns is the number of scatterers and is a parameter that

depends on the geometry of the problem.
Convergence to the Rayleigh approximation can be assessed
according to the amplitude SNR defined in (1). Following [22],
the K-distribution moments are given by
 


 2  /2 1 + 2 + 2

. (4)
m = E {A } = 2
/2 ()
Inserting (4) to (1), one obtains the amplitude SNR for the
K-distribution
1
. (5)
r=



 
( ()) / 1 + 12 + 12 1
It should be noted that according to (4), the expected value
of the intensity signal m2 is equal to 2 2 . Thus, the linear correspondence between the ultrasound mean intensity and UCA
concentration, even at low concentrations, can be explained by
the K-distribution model, since 2 is proportional to the number
of scatterers in the resolution cell [23].
B. Model for Speckled CEUS Images
There is no unanimously accepted model for the formation
of speckled CEUS images; therefore, the multiplicative speckle
noise model, widely used for the denoising of B-mode images,
was adopted here. A general multiplicative model for a speckle
noise was proposed in [5]
g [n, m] = f [n, m] u [n, m] + [n, m]

(6)

1971

where g and f are the measured and the original image, respectively. The multiplicative component of the noise is given by u,
represents the additive noise component. Here, n and m are the
radial and angular coordinates, respectively, for a phased-array
transducer, or axial and lateral coordinates for a linear transducer. The multiplicative component of the noise relates to the
physics of coherent imaging, the additive noise component represents electrical measurement noise and phenomena not fully
described by the model.
In [8], it was claimed that compared to the stronger specklerelated multiplicative noise component, the additive noise in
B-mode ultrasound images is negligible. The same strategy is
used here in the context of DCEUS scans despite the lower SNR
of the harmonic signal. Applying the log transform to (6), the
multiplicative speckle noise u is converted into an additive term.
Defining g, f, and u
as the logarithms of g, f, and u, respectively,
one obtains
g [n, m, l] = f [n, m, l] + u
[n, m, l]

(7)

where l is used as the frame index. The resulting additive noise

u
has very different correlation properties in the spatial and
temporal domains. In the time domain, noise in consecutive
frames is expected to be independent due to the movement of gas
bubbles over time. In the spatial domain, the log-transformed
noise was proven to have a significant structure [8]. Consequently, temporal denoising of DCEUS cines is easier to implement, without the need for preliminary decorrelation of the
signal.
In the search for a suitable denoising algorithm, the first-order
statistics of the log-transformed noise should be considered. For
high enough concentrations of UCA, and assuming the amplitude of CEUS signals follows the Rayleigh distribution, the
log-transformed signal is known to obey the FisherTippet distribution [24]




pg (x) = 2 exp 2x ln 2 2 exp 2x ln 2 2 . (8)
This distribution has a double exponent formation. It was
shown in [24] that if the first three terms of the Taylor series
of the second exponential is inserted into (8), a Gaussian approximation for (8) is obtained. This result is helpful since it
presents a Gaussian approximation to the FisherTippet distribution, with a
constant variance var{
g } = 0.25 and an expected
value = ln( 2), which is a function of the scatterers concentration. This independence between the values of the variance and the expected value enables the representation of g as a
sum of f, a signal representing the UCAs concentration, and a
noise u
, with constant variance and = 0. This approximation
is fairly accurate when x is close to its expected value , while
at the tails of the FisherTippet distribution the approximation
error increases considerably [24]. Due to the complexity of the
K-distribution, a similar analytical approximation could not be
derived for low UCA concentrations. However, after log transform, signals obeying the K-distribution display a heavy tail
similar to those characterizing the FisherTippet distribution. In
accordance, the denoising procedure was developed to suit the
FisherTippet distribution. The denoising procedure was later
validated and shown to be effective (see below) over a wide range
of UCA concentrations including low UCA concentrations

1972

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO. 8, AUGUST 2015

for which the noise does not fit the FisherTippet model and
can be described using the K-distribution.

III. DENOISING OF DCEUS CINE SEQUENCES

C. Dynamics of CEUS Cines

Since this study focuses on DCEUS data processing in the
time domain, a model for the dynamics of CEUS cines must
be derived. The degree to which this model can account for the
underlying physical phenomena will have a direct effect on the
reliability and accuracy of the results.
In most quantitative DCEUS studies, an indicator-dilution
model is fitted to the CEUS signal over time, facilitating
estimation of physiological meaningful parameters from the
time-intensity curve (TIC). A review of five indicator-dilution
models and their properties can be found in [25]. The two models that best fit the examined in vivo data were the log-normal
and diffusion-with-drift models. In order to increase SNR and
facilitate the parameter estimation process, the signal is typically
first averaged over a predefined spatial region of interest (ROI)
around each pixel. However, both the initial spatial averaging
and the use of a global temporal model limit the resolution of the
analysis, and can potentially mask important local information,
in both time and space. In addition, although indicator-dilution
models were shown to fit DCEUS signals measured for homogeneous, well perfused areas, they only match the first pass of the
UCA and, thus, are problematic as a mean for signal denoising.
The log-normal model is the most widely used model for
DCEUS data analysis. The log-normal distribution function is
given by
AUC
2
2
e([ln(tt 0 )] /(2 )) + I0 ,
I (t) =
2 (t t0 )

t > t0

(9)
where AUC is the area under curve of the fitted model, t0 is
the time delay before the beginning of the wash-in phase, I0
is the intensity offset. On a logarithmic time scale, and
are the location and scale parameters, respectively. Due to its
popularity, this model is an important reference used for the
evaluation of new DCEUS processing methods.
The diffusion-with-drift model is based on a fluid-dynamic
interpretation of the indicator-dilution process. The combination
of diffusive and convective transport is described by
2 C (x, t)
C (x, t)
C (x, t)
=D
v
t
x2
x

model UCA concentration signals are expected to be smooth,

but can include isolated sharp transitions.

(10)

where C(x, t) is the indicators concentration at a given time

t and location x. Here, v is the blood velocity and D is the effective diffusion constant. In the case of constant velocity, the
diffusion term results in the indicators concentration becoming
increasingly smooth over time as the UCA bolus flow through
the vasculature. This assumption holds for small systemic vessels and for the microvasculature. However, pulsatile flow
in large vessels can result in sudden indicator concentration
changes. In addition, tumor vasculature is highly heterogeneous
and irregular, resulting in extremely long circulation times;
thus, making the distinction between the first passage of UCAs
and following recirculation difficult. Since a piecewise-smooth
model for UCA concentration can cope with all mentioned phenomena, it will be used in this study. Specifically, with this

A. Linear Versus Nonlinear Filtering

As the processing of the DCEUS measurements is performed
for each pixel separately, the spatial coordinates will be used
here as indices. For reasons of clarity, these indices will be
omitted for the rest of this section and the notation g[l] = f[l] +
u
[l] will be used, replacing the notation used in (7).
Temporal compounding is equivalent to the application of a
specific FIR filter. The definition of this filter depends on the
number of temporal samples being compounded
1
f [l] =
M
=

(M 1)/2

g [l i]

i=(M 1)/2

N
1


[k] H [k] exp
G

k =0

i2kl
N


(11)

where M is the number of samples of the moving window, f is

is the Fourier transform of g[l]. The
the estimation of f, G[k]
Fourier transform of the rectangular moving average filter is
given by H[k]. Different finite-impulse response filters may be
designed according to the desired tradeoff between the cut-off
frequency, roll-off, ripple level, and the order of the filter. The
application of a linear filter requires the setting of a single cutoff frequency separating the slow-changing UCA concentration
dynamics and the wide-band noise. The value of this cut-off
frequency is assumed to be valid for the whole dataset and
should be set a priori. In practice, such a priori knowledge of
the cut-off frequency is generally unavailable.
Linear estimation of the temporal UCA concentration signal
may be performed on the signals representation in any given
basis. Given, a basis a general linear estimation of the concentration signal is described by
f =

N
1

a [k] 
g , [k] [k]

(12)

k =0

where is the biorthogonal basis. Here again, the values of the

coefficients a[k] have to be determined in advance. Piecewisesmooth signals are known to have sparse representations in
wavelet bases. Since the smoothness level of DCEUS signals
is generally unknown and sharp transitions can be found in
many cases during the wash-in phase, a wavelet basis is used
here for the representation of DCEUS signals in the temporal
domain. The sparse representation of UCA concentration curves
in wavelet bases is used here as a prior for signal denoising.
Wavelet denoising by soft thresholding is a nonlinear estimation method [17], which filters out the noise by adaptively
selecting the relevant subset of the basis according to their relative value. Following [26], the formulation of this denoising
operation is given by
f =

N
1

k =0

ak (
g , [k]) 
g , [k] [k].

(13)

BAR-ZION et al.: DENOISING OF CONTRAST-ENHANCED ULTRASOUND CINE SEQUENCES BASED ON A MULTIPLICATIVE MODEL

When given the threshold T, the soft attenuation rule is

defined by


T
, 0 1.
(14)
0 ak (x) = max 1
|x|
The threshold
 proposed in [17] was the universal threshold
rule T = (2ln(N )), where N is the length of the signal
and is estimated as the median of the absolute value of the
wavelet coefficients in the first decomposition level. Although
lower thresholds were presented, for example, by [18], the universal threshold rule was used in this study, due to its robustness
to high levels of noise. When this value of T is selected, the
wavelet thresholding algorithm produces a nearly minimax risk
for piecewise regular signals, which means that this estimator is
close to the optimal mean-square error for every signal in this
class [17].
The overall scheme proposed in [17] can be summarized by
the three-step wavelet shrinkage algorithm. First, the empirical wavelet coefficients are calculated from the measurements,
using the wavelet transform. Then, the soft threshold rule is applied to the wavelet coefficients, decreasing the amplitudes of
all coefficients and excluding those associated with the noise.
Finally, the inverse wavelet transform is applied, producing an
estimation of the clean DCEUS signal. Even though this threshold rule excludes coefficients produced by the Gaussian noise
with high probability, it is very sensitive to noise outliers. The
outlier-resistant wavelet denoising algorithm and the consideration for selecting a specific wavelet basis are described below.
B. Outlier Resistant Wavelet Denoising
The wavelet denoising algorithm presented in [17] and [18]
was optimized for the removal of the Gaussian white noise.
While the noise in consecutive DCEUS frames can be assumed
to be independent, its distribution is heavy tailed as described in
previous sections. The three-step wavelet denoising algorithm is
sensitive to noise outliers, since their projections on the wavelet
basis produce large coefficients that are preserved as features
of the signal. The outlier resistant wavelet denoising scheme
copes with the non-Gaussian noise distributions by introducing
an outlier detection step and removing these outliers from the
signal before each step of the wavelet decomposition [19].
Following the notations of [24], for every level of decomposition (k), a smooth version of the detail coefficients gk , is
calculated. This smooth signal is calculated using a moving median filter with a window of seven samples. Subsequently, the
robust residuals of level k, Rk are calculated according to the
following rule:

0,
|gk | < k
Rk [l] =
sign (gk ) (|gk | k ) , |gk | k
gk gk [l] gk [l]

(15)

where k is the threshold for level k, defining a small percentage p of the coefficient gk as outliers. Next, the detected
residuals can be subtracted from gk and the next level of the
wavelet decomposition can be performed. As discussed in [19],
short decomposition wavelet filters reduce the leakage of outlier

1973

patches into the smooth coefficients, isolating them and facilitating their removal. On the other hand, long reconstruction filters
are desirable in order to ensure a sufficient level of smoothness.
Following [19] the b-spline biorthogonal wavelets with three
and nine vanishing moments were used in this study.
In this study, the noise contaminating the DCEUS movies
contains isolated outliers in the time domain. In [24], the Fisher
Tippet noise containing isolated outliers was processed by a
single use of the outlier removal procedure. It was shown that
this outlier removal step successfully removed the heavy tail
of the distribution, producing the semi-Gaussian noise. At the
same time, the short moving median had a negligible effect
on the signal. Similarly, a single use of the outlier removal
procedure was used in the current study. The resulting semiGaussian noise can then be removed using the standard threestep algorithm with the universal threshold rule. This processing
scheme is robust to the noise outliers and also incorporates the
numerical efficiency of wavelet denoising.
IV. VALIDATION OF THE PROPOSED MODEL
All the phantom and in vivo experiments were performed using the Vevo 2100 (VisualSonics Inc., Toronto, Canada) scanner
and the linear array MS250 transducer in contrast mode. Amplitude modulation is used in the Vevo2100 for the separation
of CEUS signal. The cines were exported from the scanner as
linear raw data files (without log compression).
A. Phantom Experiments
In order to evaluate the performance of the proposed denoising algorithm when compared to established methods, a series
of flow phantom experiments were performed.
1) Ultrasound Contrast Agents: MicroMarker UCA was used
in all the phantom and in vivo experiments. Before each experiment, a new vial was activated according to the companys
protocol. Following the activation, the size distribution of the
microbubbles was measured using a Beckman Coulter Multisizer 3 system. In order to obtain reliable quantification, 10 L
of UCA was diluted in 10 mL of Isoton II diluent [9]. In each
test, 100 L of this solution was drawn by the Multisizer and
each measurement was repeated three times. The aperture used
detects bubbles with radii as small as 0.7 m.
2) Continuous-Flow Phantom Setup: The purpose of the
continuous-flow phantom experiments is threefold: to detect the
linear region of the relationship between the UCA concentration
and the ultrasound intensity, to validate the K-distribution
model, and to compare the effectiveness of the MIP, temporal
compounding, and wavelet denoising algorithms using movies
containing homogenous concentrations of UCA. The setup of
this phantom is described in Fig. 1(a). Before each injection, a
solution of deionized water and UCA was mixed using a magnetic stirrer. Then, the solution was loaded into a 3-mL syringe
and injected using a syringe pump (model NE-4000, New Era
Pump Systems, Inc., Farmingdale, NY, USA) into a polyethylene tube (internal diameter of 1.4 mm) submerged inside a small
water tank. The injection rate was set to 7.7 L/s. Cines of 300
frames at 25 frames/s were recorded. The scans were repeated
using four different vials of MicroMarker. The transducer was

1974

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO. 8, AUGUST 2015

TABLE I
CONTRAST AGENTS CONCENTRATIONS USED IN CONTINUOUS-FLOW
PHANTOM EXPERIMENT
Relative
concentration

A
A/2
A/5
A/10
A/20
A/50
A/100
A/200
A/500
A/1000
A/2000
A/5000

Contrast agent
concentration
(%)

Scatterers per
ultrasound resolution
cell

Bubbles per unit

volume
(bubbles/L)

5%
2.5%
1%
0.5%
0.25%
0.1%
0.05%
0.025%
0.01%
0.005%
0.0025%
0.001%

743.5
371.7
185.9
74.35
37.17
18.59
7.44
3.72
1.86
0.74

67500 3500
33750 1750
13500 700
6750 350
3375 175
1350 70
675 35
337.5 17.5
135 7
67.5 3.5
33.75 1.75
13.5 0.7

The number of effective Scatterers per ultrasound resolution cell was estimated by fitting
the measured amplitude SNR to the K-distribution model (see Fig. 2).

Fig. 1. Phantom setup. (a) Continuous-flow setup during the scanning phase.
(b) Indicator-dilution setup.

aligned to the center of the flow channel with an in-plane angle of about 30 to minimize reverberations. A syringe pump
was used in this experiment in order to reduce flow velocity
fluctuations.
Several measures were made to improve the quality of the
measurements. Before mixing each solution, the vial of the
contrast agents was shaken to achieve stable microbubble concentration over time. Before each injection, deionized water was
pumped through the tubing using a peristaltic pump (Minipuls 3,
Gilson Inc., Middleton, WI, USA) until no traces of UCA
were seen by the scanner. In addition, increasing concentrations of UCA were injected in order to reduce the influence
of residual microbubbles between consecutive injections. Finally, great care was taken to remove static air bubbles from the
tubing.
For each vial, a range of contrast agent dilutions spanning
four orders of magnitude were scanned (see Table I). This range
of dilutions includes both tens of microbubbles in each resolution cell, complying with the Rayleigh distribution approximation, and resolution cells containing single bubbles. Assuming
that the average blood volume of a laboratory mouse is about
1.52.5 mL; a 50 L bolus injection of UCA would results in a
preclinical concentration of around 2.5%.
The recorded cines were loaded into a personal computer
and analyzed using dedicated MATLAB code (The Mathworks,
Natic, MA, USA). A ROI was selected in each cine, containing
a cross section of the tube with a negligible signal originating
from the walls of the tube. Since three tests were performed
using this phantom setup, three different processing methods
were applied to the data. First, in order to detect the linear region of the UCA concentration versus ultrasound intensity, the
intensity of the pixels inside the ROIs was averaged throughout the movie. Next, in order to validate the K-distribution
model, the amplitude SNR was calculated inside the ROI for
every frame and averaged throughout the cine. Then, the SNR
values were fitted to the K-distribution SNR model of (5).

The K-distribution model enabled the normalization of the concentration axis so that the SNR can be presented as a function of
the number of effective scatterers per resolution cell. Finally, the
log-transformed cines which were processed using the different
algorithms were compared in order to evaluate the effectiveness
of the algorithms in estimating the TIC and the noise removal.
The moving average filter of the temporal compounding algorithm was chosen to be 21 samples long, to achieve a reasonable
noise reduction, while avoiding oversmoothing the data during
the wash-in phase. Following the considerations described in the
previous section, wavelet denoising was performed using the 3,
9 biorthogonal wavelets. The robust residual removal procedure
was set so that the length of the moving-median window was 7
and p was set to 7%.
The spatial coefficient of variation was used to evaluate the
ability of the different algorithms to remove the speckle noise.
The underlying assumption was that the intensity within a small
ROI should be uniform. The spatial coefficient of variation (CV )
was defined as the ratio of the standard deviation to the mean
of log-intensity value in each processed movie within the
ROI.
3) Indicator-Dilution Phantom Setup: An indicator-dilution
phantom was used to test the effectiveness of the different algorithms on real scans containing changing concentrations of
UCAs. The setup of this phantom, based on [27], is described in
Fig. 1(b). The three-way stopcock was used to connect both the
reservoir and the syringe pump to the phantom. The peristaltic
pump was used to maintain a steady flow of 36 L/s within the
system. The syringe pump, filled with undiluted UCA, was used
to inject a bolus of 50 L into the phantom. Cines of 920 frames
at 25 frames/s were recorded using the Vevo 2100 scanner in
contrast mode. The alignment of the MS250 probe was carried
out as described in the previous section.
The recorded cine was loaded into a personal computer and
analyzed using MATLAB. A ROI was selected as described
above. Regional intensity measurements were calculated by averaging the intensity of the pixels included inside the entire
5 5 pixel ROI. Local TIC were estimated from single pixels.

BAR-ZION et al.: DENOISING OF CONTRAST-ENHANCED ULTRASOUND CINE SEQUENCES BASED ON A MULTIPLICATIVE MODEL

1975

Fig. 2. Continuous-flow phantom results. (a) Linearity of the relationship between the raw data intensity and UCA concentration. (b) Measured amplitude SNR
and its compliance to the K-distribution model. (c) Spatial coefficient of variation for different processing durations (0.1% UCA). (d) Spatial coefficient of variation
for different concentrations of UCA.

The MSE and correlation coefficient (R) criteria were used to

evaluate the discrepancy between the regional measurements
and local TIC estimations.
B. Processing of In Vivo Data
1) Ex-Utero Mouse Embryos: In order to evaluate how the
proposed algorithm functions when a healthy well-organized
vasculature is imaged, a subset of the mouse embryo scans
studied in [28] was reprocessed. Specifically, the scans used
were of embryos imaged at the embryonic day (E) 17.5 using
untargeted MicroMarker at a frame rate of 1 frames/s. This low
frame rate was used in order to capture the full duration of the
wash-out phase. Full details about the experimental preparation,
microbubble reconstruction, injection procedure, and ultrasound
imaging can be found in [28].
2) Human LS174T and PC3 Xenografts in Mice: Xenograft
tumors were induced in mice using either LS174T human

colorectal cancer cells or PC3 human prostate cancer cells. Both

cell lines were cultured in Gibco DMEM (Life Technologies
Inc., ON, Canada) supplemented with 10% fetal bovine serum
(Hyclone, UT, USA). One million cells were injected subcutaneously into the left hind limb of each female eight-week old
SCID nude mouse (Charles River, CA, USA). The tumors were
scanned after they reached a target size of 46-mm depth. All
procedures were completed with the animal anesthetized under
isoflurane and in accordance with Sunnybrook Health Science
Centres approved protocol for animal care and use.
The transducers field of view was aligned to the center of
the tumor with the largest cross section. The acoustic focus was
placed around the level of the tumors center. The contrast gain
was set to 32 dB and the frame rate to 25 frames/sec.
3) Processing of DCEUS Cines: The recorded cines
were loaded into a personal computer and analyzed using
MATLAB. In order to evaluate the effectiveness of the different processing algorithms, the log-transformed processed cines

1976

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO. 8, AUGUST 2015

were compared. The parameters used were those described in

the phantom experiments section.
For quantitative assessment of the different processing methods, two measures were used: spatial coefficient of variation
and spatial autocorrelation. In order to evaluate the ability of
the different algorithms to remove the speckle noise from the
ex-utero scans, the spatial coefficient of variation was measured
within a ROI placed inside the carotid arteries of the mouse
embryos. The carotid arteries are big vessels in which homogeneous distribution of UCA may be assumed. Differences in
spatial coefficient of variation values across the different methods were tested using two-sample t-tests. A two-sided p-value
<0.05 was considered to be statistically significant. The other
measure was the support of the autocorrelation function exceeding 90% of the maximal value, in pixels. This measure was used
in ultrasound image processing studies to compare the resolution of different images [8].
V. RESULTS
A. Phantom Experiments
Fig. 2 presents the results of the experiments with the
continuous-flow phantom. The results are separated into three
categories: a calibration plot, evaluating the linear correlation
between the mean intensity of the raw data and the concentration of microbubbles; validation of the K-distribution model
according to the SNR measured at different concentrations; and
characterization of the spatial coefficient of variation over time,
and over a wide range of concentrations. The two highest concentrations were not included in the analysis since a significant
shadowing effect was observed in all their scans.
The mean intensity measured as a function of UCA concentration in the phantom, is presented in Fig. 2(a). The values represent the mean intensity averaged over all experiments,
and the error bars represent the standard deviation measured
when using UCA from the different vials. The results show
a linear relationship between the UCA concentrations and the
mean intensity measures which holds for 2 orders of magnitude. The linearity tails off for both the highest and the lowest
concentrations.
Fig. 2(b) shows a side-by-side comparison between the calculated SNR value in the ROI and that from the fitted K-distribution
SNR model. The SNR model was used to normalize the UCA
concentrations so that the measurements are presented as a function of the effective scatterers per resolution cell. The SNR increases monotonically until reaching a plateau at a value of
1.91. At the lowest concentration, an average of 0.7 effective
scatterers may be found in a resolution cell in each frame.
The spatial coefficient of variation (Cv ) is presented in
Fig. 2(c) when measured inside the ROI as a function of time, at
a concentration of 0.1%. While TC and WD produce fairly low
Cv over the whole scan, the Cv produced by MIP is markedly
higher at the beginning of the cine. The Cv value of the MIP at t0
is 0.52, the value expected from a single speckled CEUS frame.
This value drops after a few seconds to a level comparable with
TC. WD shows the lowest Cv throughout the cine. Similar results were recorded for different UCA concentrations. Fig. 2(d)
shows Cv averaged over time, as a function of concentration.
The Cv values were averaged over the first 300 frames. The Cv

Fig. 3. Continuous-flow phantom results. Single frames taken from the different processed cines (t = 1 s).

produced by WD is the lowest between the three algorithms,

while that produced by MIP is the highest. A reduction of Cv
values at low UCA concentrations was recorded for both TC
and WD.
The noise reduction performances of the different algorithms
are visualized in Fig. 3. This figure includes a single frame from
each processed movie, taken one second after the beginning of
the scan. These images show how well the homogenous UCA
concentration within the center of the phantom correlates with
the intensity in the processed CEUS movies. The frame taken
from the WD movie is less noisy and more homogenous both
within the lumen of the phantom and in the background.
Fig. 4 presents the results of experiments performed with
the indicator-dilution flow phantom. The temporal denoising
performance of TC and WD are compared against the MIP
and the log-normal model fitting. Both local TIC estimation
[single pixel, Fig. 4(b)] and regional TIC estimation [5 5 pixel
averaged, Fig. 4(a)] are presented. While TC, MIP, and the
model fitting results change when local estimation is attempted,
regional and local WD results are very similar.
Numerical analysis of local TIC estimation is presented in
Table II. The local TICs were estimated from single pixels within
a 5 5 ROI. The averaged intensity within this ROI in each
frame was used as a point of reference, when evaluating the
quality of local TIC estimation. The quality of the estimation was
quantified by the MSE and correlation coefficient (R) between
these estimations and the averaged regional measurements. The
proposed method produced both the highest R values and lowest
MSE values of all the compared methods.
B. In Vivo Data
An example of a nonprocessed DCEUS scan of a mouse embryo is shown on the left row of Fig. 5. Three frames taken

BAR-ZION et al.: DENOISING OF CONTRAST-ENHANCED ULTRASOUND CINE SEQUENCES BASED ON A MULTIPLICATIVE MODEL

1977

TABLE II
MSE AND CORRELATION COEFFICIENTS OF SINGLE PIXEL TIC ESTIMATIONS

Fitting
MIP
TC
WD

MSE [a.u.]

0.707 0.004
0.641 0.011
0.719 0.009
0.742 0.003

9.70 2.00
172.34 21.44
7.46 1.73
6.93 1.48

of TC and WD to movement, compared to MIP. Although the

hind-limb of the mouse is held in place, the cine contains small
motion artifacts that are accumulated by the MIP and create a
blurring effect. This effect can be seen clearly by comparing
the size of the air bubble trapped above the tumor. The bubble in the MIP frames appears considerably larger than in the
corresponding TC and WD frames.
The quantitative results acquired from the in vivo data are
presented in Fig. 7 and Table III. The proposed method achieved
significant improvement in the spatial coefficient of variation,
measured within homogeneous regions (9 9 ROI inside the
carotid arteries of mouse embryos, Fig. 7) in comparison to both
MIP and TC. In addition, comparing the support of the axial and
lateral autocorrelation function, a considerable improvement in
resolution was measured in the processed WD scans, compared
to the corresponding MIP frames (see Table III).

VI. DISCUSSION

Fig. 4. Indictor-dilution phantom results. (a) Regional TIC (5 5 averaged

pixels). (b) Local TIC from a single pixel.

during the wash-in phase (50 s), the peak-enhancement (100 s),
and the wash-out phase (250 s) are presented on the top, middle, and bottom column. To the right, the corresponding frames
from the processed MIP, TC, and WD movies are presented,
allowing us to compare the three methods. Since MIP holds on
to the maximum intensity value over time, there is no significant
change between the MIP image at the peak-enhancement time
and wash-out time. Moreover, the high concentration of UCA
entering the liver during the first few frames of the wash-in is
retained by MIP throughout the movie. In contrast, the UCA
signal decrease can be appreciated in both TC and WD images
(for example, inside the brain). The ability of the WD to reject noise from DCEUS images is noticeable when observing
homogeneous areas such as the carotid arteries.
The shapes of single blood vessels in the core of a hind-limb
tumor depicted by WD are presented in Fig. 6. Fig. 1ex (supplementary material) compares the WD results with MIP and
TC. An interesting feature observed in Fig. 1ex is the robustness

While in most studies isotropic spatial filtering is performed

before TIC estimation, it was shown here that using the proposed
denoising method, TIC could be estimated locally from single pixels. When tested on the indicator-dilution phantom data,
the single pixel estimations provided good agreement with the
regional TIC estimation and correlated well with the instantaneous intensity measurements averaged over the ROI. The main
limitation of the standard isotropic averaging is the resulting
reduced spatial resolution. In addition, speckle noise contains
significant spatial structure, thus independence between noise
measurements in adjacent pixels cannot be assumed. The denoising was performed here only in the temporal domain, and
was able to remove the speckle noise and produce clear images
that describe the change in UCA concentration over time. As
depicted in Fig. 3, the WD algorithm provides the homogenous
concentration of UCA within the phantom, while removing measurement noise from the surrounding regions. Fig. 5 presents a
clear depiction of the UCA concentration in regular vasculature,
while Fig. 6 shows the vasculature in a tumor. The tumor scans
are characterized by lower UCA concentrations and demonstrate
the ability of WD to produce clear depictions of isolated blood
vessels in the tumors core.
The linear correspondence between the UCA concentration
and the intensity of the ultrasound echoes is the basis of quantitative contrast-enhanced ultrasound imaging. Fig. 2(a) shows
linear correlation between UCA concentration and Intensity,
over 2 orders of magnitude. Deviation from linearity observed at
concentration extremes can be attributed to acoustic shadowing
effects at high concentrations, and dominant tissue (phantom)
harmonics at low concentrations.

1978

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO. 8, AUGUST 2015

Fig. 5. Single frames from different processed cines during wash-in, peak enhancement, and wash-out phases of a mouse embryo scan. The noise levels are
considerably lower in WD images especially during the wash-in phase (see arrows), while MIP presents partial filling of blood vessels. False structures in the
MIP scans and their locations in WD images are marked by stealth arrows. The ability of WD to present the wash-out phase is easily seen in the brain tissue
(arrowheads).

Fig. 6.

Single frame from the wash-in phase in WD cine of LS174T tumor.

The models proposed in this study were supported by the

experimental results. The SNR values measured using the
continuous-flow phantom corroborate the K-distribution SNR
model [see Fig. 2(b)]. The favorable results achieved by both
TC and WD in this paper support the validity of the multiplicative noise model for CEUS images. Therefore, the proposed
models can serve as a basis for applying different signal processing methods to DCEUS data. These models are general and
should hold for both clinical and preclinical scans.
Denoising of DCEUS signals enables estimation of UCA
concentrations during both wash-in and wash-out phases. Some
variants of MIP enable the depiction of the wash-out phase by
attaching a decaying weight to the measurements. However, this
is an artificial modification that does not produce a direct relationship between the UCA concentration during the wash-out
phase and the MIP values. The growing research in the area of
targeted drug/gene delivery may benefit most from quantitative
imaging of targeted CEUS and, thus, from the denoised washout data, where possible new applications may arise once this
kind of data is available.

Fig. 7. Spatial coefficient of variation measured inside the carotids of the

mouse embryos in different processed cines.

It was also found that in MIP movies the Cv values are initially
high, even when the concentration of UCA is constant. The Cv
values slowly decrease only after few seconds [see Fig. 2(c)]
to a level comparable to TC and WD. This means that the MIP
cines are very noisy at first and gradually become clearer. This
delayed response is problematic when the dynamics of DCEUS
cines is studied.
The proposed method has two main theoretical advantages
over classical TC. The first is that TC relies on the implicit assumption of the Gaussian noise. However, it is well known that
log-transformed CEUS signals suffer from low-value outliers.
These outliers motivated the introduction of a preprocessing
step for their removal. The second advantage of WD is that unlike TC, it does not require the definition of a cutoff frequency
a priori. Moreover, a low-pass filter cannot separate the signal
and the noise efficiently since they have an overlapping supports in the frequency domain. Therefore, the optimal cut-off
frequency depends on the local flow characteristics and is difficult to estimate in advance. In contrast, WD relies on the sparse

BAR-ZION et al.: DENOISING OF CONTRAST-ENHANCED ULTRASOUND CINE SEQUENCES BASED ON A MULTIPLICATIVE MODEL

TABLE III
RESOLUTION RATIO GAIN OF WD IN COMPARISON TO MIP

LS174T tumors
PC3 tumors
Mouse Embryos
All scans (n = 20)

Axial Resolution Ratio

Lateral Resolution Ratio

1.66 0.38
1.96 0.26
2.27 0.19
1.99 0.37

1.45 0.1
1.74 0.05
1.62 0.13
1.60 0.15

representation of piecewise-smooth functions in wavelet bases

and adaptively removes the noise from the data. WD retains the
dominant fine details of the signal, while efficiently smoothing
out other regions in which the signal changes slowly. This can be
considered as setting different cut-off frequencies for different
temporal segments. Thus, while both methods are parameter dependent, TC requires adjusting the upper frequency limit, while
for WD the noise level is estimated from the measured signals.
These theoretical advantages correspond to the improved denoising capabilities of WD, for a wide range of UCA concentrations in the in vitro experiments [see Fig. 2(d)]. Corresponding
quantitative in vivo results were presented in Fig. 7.
Indicator-dilution models enable the reliable analysis of
CEUS signals only when the model fits the dynamics of the underlying UCA concentrations and the noise level is sufficiently
low to enable the estimation of the models parameters. When
the underlying signals do not correspond to the fitted model
or the noise level is high, the estimation is unsatisfactory (see
Fig. 4). Denoising is an important preprocessing procedure for
indicator-dilution model fitting, enabling the noise removal and
robust TIC estimation, while including a high level of flexibility.
Denoising is also useful if the effects of UCA recirculation are
dominant and the first passage of UCA should be detected and
analyzed separately.
Denoising of DCEUS signals using WD and TC improved the
spatial resolution in comparison to MIP analysis; MIP retains
the maximal intensity over time, thus movement blurs the image
(see Table III). Processing schemes of scans containing significant motion artefacts should incorporate motion-compensation
algorithms, but the residual movements are expected to be better
handled using denoising methods.
There are two main limitations of the proposed methods in
comparison to MIP. The first is the ability of MIP to capture
the flow trajectories of extremely sparse microbubbles. MIP is
nonlinear and is very sensitive to single bubbles. However, the
outlier removal rule can be adjusted to discriminate between
low-value noise-related outliers and positive outliers produced
by single bubbles. This is a subject of future research. The second limitation is that MIP retains the positive outliers of the
intensity distribution and by that improves the contrast to tissue
ratio. This characteristic can be important in preclinical systems, having low contrast to tissue ratio. Adjusting the contrast
specific pulse sequences and the dynamic range of the scanner
to minimize tissue harmonics is important in these cases.
VII. CONCLUSION
To conclude, temporal denoising of log-transformed DCEUS
data was shown to be an effective framework for processing

1979

DCEUS signals: WD produces superior results in reducing the

variability of DCEUS signals and removing speckle noise from
these cines. WD enabled the estimation of local TIC without the
need for spatial isotropic filtering. A substantial improvement
in noise reduction was demonstrated over a wide range of UCA
concentrations. The increased SNR resulted in a significant improvement in the depiction of homogenous regions (p < 0.01).
Denoising of DCEUS signals enables the estimation of UCA
concentration maps through the duration of the scan, revealing new observations on CEUS dynamics. Finally, denoising of
DCEUS signals was proved to be relatively resistant to motion
artifacts, having higher resolution by a factor of 2 in the axial
direction and 1.6 in the lateral direction, as compared to MIP.
ACKNOWLEDGMENT
The authors would like to thank J. Denbeigh from the Department of Medical Biophysics at the University of Toronto for
providing mouse embryo scans used in this study. F. S. Foster
discloses that he is a Consultant to VisualSonics.
REFERENCES
[1] B. B. Goldberg et al., Ultrasound contrast agents: A review, Ultrasound
Med. Biol., vol. 20, no. 4, pp. 319333, 1994.
[2] S. Qin et al., Ultrasound contrast microbubbles in imaging and therapy:
Physical principles and engineering, Phys. Med. Biol., vol. 54, no. 6,
pp. R27R57, Mar. 2009.
[3] M. Lampaskis, Investigation of the relationship of nonlinear backscattered ultrasound intensity with microbubble concentration at low MI,
Ultrasound Med. Biol., vol. 36, no. 2, pp. 306312, Feb. 2010.
[4] R. J. Eckersley et al., Optimising phase and amplitude modulation
schemes for imaging microbubble contrast agents at low acoustic power,
Ultrasound Med. Biol., vol. 31, no. 2, pp. 213219, Feb. 2005.
[5] A. K. Jain, Image filtering and restoration, in Fundamentals of Digital
Image Processing, vol. 3. Englewood Cliffs, NJ, USA: Prentice-Hall,
1989, pp. 267341.
[6] X. Zong et al., Speckle reduction and contrast enhancement of echocardiograms via multiscale nonlinear processing, IEEE Trans. Med. Imag.,
vol. 17, no. 4, pp. 532540, Aug. 1998.
[7] X. Hao et al., A novel multiscale nonlinear thresholding method for
ultrasonic speckle suppressing, IEEE Trans. Med. Imag., vol. 18, no. 9,
pp. 787794, Sep. 1999.
[8] O. V. Michailovich and A. Tannenbaum, Despeckling of medical ultrasound images, IEEE Trans. Ultrason. Ferroelectr., Freq. Control, vol. 53,
no. 1, pp. 6478, Jan. 2006.
[9] A. R. Sureshkumar, Investigation into the origin and nature of variability
in quantitative measurements of tumour blood flow with contrast-enhanced
ultrasound, M.Sc. thesis, Dept. Med. Biophys., Univ. Toronto, Toronto,
ON, Canada, 2012.
[10] M. P. J. Kuenen et al., Maximum-likelihood estimation for indicator
dilution analysis, IEEE Trans. Biomed. Eng., vol. 61, no. 3, pp. 821831,
Mar. 2014.
[11] G. Barrois et al., A multiplicative model for improving microvascular flow estimation in dynamic contrast-enhanced ultrasound (DCE-US):
Theory and experimental validation, IEEE Trans. Ultrason. Ferroelectr.,
Freq. Control, vol. 60, no. 11, pp. 22842294, Nov. 2013.
[12] S. R. Wilson and P. N. Burns, Microbubble-enhanced US in body imaging: What role?1 , Radiology, vol. 257, no. 1, pp. 2439, Oct. 2010.
[13] R. A. Linden et al., Contrast enhanced ultrasound flash replenishment
method for directed prostate biopsies, J. Urol., vol. 178, no. 6, pp. 2354
2358, Dec. 2007.
[14] J. K. Dave et al., Static and dynamic cumulative maximum intensity
display mode for subharmonic breast imaging a comparative study with
mammographic and conventional ultrasound techniques, J. Ultrasound
Med., vol. 29, no. 8, pp. 11771185, Aug. 2010.
[15] M. A. Pysz et al., Assessment and monitoring tumor vascularity with
contrast-enhanced ultrasound maximum intensity persistence imaging:,
Invest. Radiol., vol. 46, no. 3, pp. 187195, Mar. 2011.

1980

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 62, NO. 8, AUGUST 2015

[16] R. C. Gessner et al., Blood vessel structural morphology derived from

3D dual-frequency ultrasound images, in Proc. IEEE Ultrason. Symp.,
2010, pp. 209212.
[17] D. L. Donoho, De-noising by soft-thresholding, IEEE Trans. Inf. Theory,
vol. 41, no. 3, pp. 613627, May 1995.
[18] D. L. Donoho and I. M. Johnstone, Adapting to unknown smoothness via
wavelet shrinkage, J. Amer. Stat. Assoc., vol. 90, no. 432, pp. 12001224,
Dec. 1995.
[19] A. G. Bruce et al., Denoising and robust nonlinear wavelet analysis,
Proc. SPIE, vol. 2242, pp. 325336, 1994.
[20] R. F. Wagner et al., Statistics of speckle in ultrasound B-scans, IEEE
Trans. Sonics Ultrason., vol. SU-30, no. 3, pp. 156163, May 1983.
[21] E. Jakeman and P. N. Pusey, A model for non-Rayleigh sea echo, IEEE
Trans. Antennas Propag., vol. AP-24, no. 6, pp. 806814, Nov. 1976.
[22] V. Dutt and J. F. Greenleaf, Speckle analysis using signal to noise ratios based on fractional order moments, Ultrason. Imag., vol. 17, no. 4,
pp. 251268, Oct. 1995.
[23] J. W. Goodman, Statistical properties of laser speckle patterns, in
Laser Speckle and Related Phenomena. Berlin, Germany: Springer, 1975,
pp. 975.
[24] O. Michailovich and D. Adam, Robust estimation of ultrasound pulses
using outlier-resistant de-noising, IEEE Trans. Med. Imag., vol. 22,
no. 3, pp. 368381, Mar. 2003.
[25] C. Strouthos et al., Indicator dilution models for the quantification of
microvascular blood flow with bolus administration of ultrasound contrast
agents, IEEE Trans. Ultrason. Ferroelectr., Freq. Control, vol. 57, no. 6,
pp. 12961310, Jun. 2010.
[26] S. G. Mallat, A Wavelet Tour of Signal Processing the Sparse Way. Amsterdam, The Netherlands: Elsevier, 2009.
[27] T. P. Gauthier, Perfusion quantification using dynamic contrast-enhanced
ultrasound: The impact of dynamic range and gain on timeintensity
curves, Ultrasonics, vol. 51, no. 1, pp. 102106, Jan. 2011.
[28] J. M. Denbeigh et al., VEGFR2-targeted molecular imaging in the mouse
embryo: An alternative to the tumor model, Ultrasound Med. Biol.,
vol. 40, pp. 389399, 2014.

Avinoam David Bar-Zion was born in Haifa, Israel,

in 1983. He received the B.Sc. degree in biomedical
engineering from the Technion, Haifa, in 2010, where
he is currently working toward the Ph.D. degree in
biomedical engineering.
He completed a year of research as a part of a collaboration between the Technion and with the Medical Biophysics Department, University of Toronto,
Toronto, ON, Canada. His research interests include contrast-enhanced ultrasound imaging, medical signal and image processing, and computer-aided
diagnosis.

Charles Tremblay-Darveau was born in St-CharlesBorromee, QC, Canada, in 1987. He received the
B.Sc. degree in physics from McGill University,
Montreal, QC, in 2009, and the M.S. degree in
medical biophysics from the University of Toronto,
Toronto, ON, Canada, in 2011 and is currently working toward the Ph.D. degree in medical biophysics at
the University of Toronto.
His research interests include nanofluidics,
Doppler contrast-enhanced ultrasound imaging, and
noninvasive blood pressure measurement with
microbubbles.

Melissa Yin was born in Nanjing, China, in 1988.

She received the B.Sc. degree in biology from McMaster University, Hamilton, ON, Canada, in 2010.
Since graduating, she has been working at Sunnybrook Research Institute, Toronto, ON, as a Research
Biologist in Imaging Research. Her research interests
include the use of contrast-enhanced ultrasound and
photoacoustic imaging to monitor disease progression and therapeutic efficacy in preclinical models of
cancer and other pathologies.
Dan Adam (M64) was born in Israel. He received
the B.Sc. and M.Sc. degrees in electrical engineering
and the D.Sc. degree in biomedical engineering from
the Technion, Haifa, Israel, in 1968, 1973, and 1977,
respectively.
He was an R&D Engineer in Biomedical Industry (Elscint Ltd.). He joined the Technion Faculty in
1977. He was with Tufts University, Boston, MA,
USA, from 1978 to 1980. He joined MIT in 1980,
teaching with the Harvard-MIT HST Program. He
rejoined the Technion in 1983, and is currently a Professor in Biomedical Engineering and was the Dean from 2009 to 2012. He
had a sabbatical at NIH, MD, from 1992 to 1993. He serves on the Board of
the Israel Society for Medical and Biological Engineering since 1988, and as
the President till 2007, and its a Delegate to the International Federation for
Medical and Biological Engineering, and the Past-Chair, Academic Division
and Council Member, European Alliance for Medical and Biological Engineering and Science. His research interests include ultrasound processingdesign
of multifrequency-phased array probes for the development of multifrequency
(spectral) imaging; ultrasound RF processing; echocardiography strain imaging, including layer-specific 2-D strain measurements as a diagnostic tool of
myocardial pathologies; ultrasound RF processing; multitransducer-phased array design; beam-forming designcoded excitation; perfusion measurements
using contrast agents; pressure estimation using contrast Agents; high-intensity
focused ultrasound; ultrasound control of thermotherapymonitoring of
thermal-cavitational therapy; cellular and subcellular processes generating arrhythmias and alternans; detection and manipulations of contrast agents; functional imaging; multimodality cardiac imaging; cardiac-MRI and echocardiography image fusion, and functional data fusion; quantification of neovasculature within plaques for estimating its vulnerability; monitoring neovasculature
within cancerous lesions; targeted drug or gene delivery using ultrasound and
microbubbles (contrast agents); noninvasive thrombolysis by focused ultrasound
in acute ischemic stroke; cardiac pacing by noninvasive high-intensity focused
ultrasound; neural activation by noninvasive high-intensity focused ultrasound.
Dr. Adam elected as the IEEE-EMBS AdCom in 1999, and to the Board of
Computers from Cardiology 1990 to 1999.
F. Stuart Foster (M90SM95) received the B.A.Sc.
degree in engineering physics from the University of
British Columbia, Vancouver, BC, Canada, in 1974,
and M.Sc. and Ph.D. degrees in medical biophysics
from the University of Toronto, Toronto, ON, Canada,
in 1977 and 1980, respectively.
He is the Founder and the former Chairman of
VisualSonics Inc., a company dedicated to preclinical imaging. Dr. Foster co-founded the Mouse Imaging Centre (MICe) now at the Toronto Centre for
Phenogenomics. He has served on the Board of Directors of the National Cancer Institute of Canada and as the Chairman of its
Committee on Research. He is currently a Senior Scientist at the Sunnybrook
Health Sciences Centre, Toronto, ON, Canada, and a Professor and the Canada
Research Chair in Ultrasound Imaging, Department of Medical Biophysics,
University of Toronto, Toronto. His current research interests include the development of high-frequency clinical and preclinical imaging systems, array
technology, intravascular imaging, photoacoustics, and molecular imaging. He
serves on numerous advisory bodies and is currently on the Editorial Boards of
Ultrasonic Imaging and Ultrasound in Medicine and Biology.
Dr. Foster is a Fellow of the American Institute of Ultrasound in Medicine.
He received the 2010 Rayleigh Award from the IEEE. He has received the Eadie
Medal for major contributions to engineering in Canada and also received the
Queens Golden Jubilee Medal, the Manning Award of Distinction for Canadian
Innovation, and the Ontario Premiers Discovery Award.