You are on page 1of 21

Physics in Medicine & Biology

You may also like


- Effective field theories of post-Newtonian
The effects of acoustic attenuation in optoacoustic gravity: a comprehensive review
Michèle Levi
signals - Modified discrete phase numerical model
for calculating ultrasonic coagulation in
Eckart flows
To cite this article: X Luís Deán-Ben et al 2011 Phys. Med. Biol. 56 6129 A V Shalunov, R N Golykh, V N Khmelev
et al.

- Reviewing the geometric Hamilton–Jacobi


theory concerning Jacobi and Leibniz
identities
View the article online for updates and enhancements. O Esen, M de León, M Lainz et al.

This content was downloaded from IP address 2.87.253.6 on 17/01/2024 at 11:14


IOP PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 56 (2011) 6129–6148 doi:10.1088/0031-9155/56/18/021

The effects of acoustic attenuation in optoacoustic


signals

X Luı́s Deán-Ben, Daniel Razansky and Vasilis Ntziachristos


Institute for Biological and Medical Imaging, Technical University of Munich and Helmholtz
Center Munich, Ingolstädter Landstraße 1, 85764 Neuherberg, Germany

E-mail: vntziachristos@gmail.com

Received 26 April 2011, in final form 18 July 2011


Published 26 August 2011
Online at stacks.iop.org/PMB/56/6129

Abstract
In this paper, it is demonstrated that the effects of acoustic attenuation
may play a significant role in establishing the quality of tomographic
optoacoustic reconstructions. Accordingly, spatially dependent reduction of
signal amplitude leads to quantification errors in the reconstructed distribution
of the optical absorption coefficient while signal broadening causes loss of
image resolution. Here we propose a correction algorithm for accounting
for attenuation effects, which is applicable in both the time and frequency
domains. It is further investigated which part of the optoacoustic signal
spectrum is practically affected by those effects in realistic imaging scenarios.
The validity and benefits of the suggested modelling and correction approaches
are experimentally validated in phantom measurements.
(Some figures in this article are in colour only in the electronic version)

1. Introduction

Optoacoustic imaging is an emerging modality with demonstrated capability of quantitative


anatomical, functional and molecular imaging within depths of several millimetres to
centimetres of tissue (Xu and Wang 2006, Razansky et al 2009, Ntziachristos and Razansky
2010). The underlying principle of optoacoustic imaging is the excitation of ultrasonic waves
due to the thermoelastic expansion caused by light absorption. Typically, a short-pulsed laser
is used as the illumination source, so that instantaneous expansion can be assumed and the
optoacoustic reconstruction consists in calculating the initial pressure (which is proportional
to the absorbed energy) from the ultrasonic signals measured at several positions outside
the imaged sample. As a result, contrast is generated due to light absorption by tissue
chromophores whereby imaging resolution is defined by the ultrasonic signal detection.
The optoacoustic image quality depends on several parameters. Of importance is the
attenuation of light propagating in tissues as a function of depth and optical properties, which
0031-9155/11/186129+20$33.00 © 2011 Institute of Physics and Engineering in Medicine Printed in the UK 6129
6130 X L Deán-Ben et al

contributes to a heterogeneous light fluence distribution and modifies the optoacoustic intensity
of the local response generated, unless corrected (Yuan et al 2007, Jetzfellner et al 2009). In
addition, acoustic heterogeneities cause time-shifting of the optoacoustic signals and/or their
distortion due to reflected or scattered waves depending on the degree of acoustic mismatch
(Xu and Wang 2003, Deán-Ben et al 2011a). To account for these image distortions, several
reconstruction approaches have been proposed to take into account the presence of acoustic
heterogeneities and reduce the respective distortion effects in the optoacoustic images (Xu and
Wang 2003, Deán-Ben et al 2011a, Zhang et al 2000, Jiang et al 2006, Cox and Treeby 2010,
Deán-Ben et al 2011c). Coupled to these effects is the acoustic attenuation and dispersion
of the propagating ultrasonic wave, which becomes significant at higher frequencies (Szabo
2004) and may distort the reconstructed images.
The ultimate penetration ability of optoacoustic imaging is defined by the combination of
optical attenuation and acoustic attenuation. Although optical attenuation is generally stronger,
the effect of acoustic attenuation can become dominant at high ultrasonic frequencies and
must also be considered. The frequency-dependent reduction of amplitude and corresponding
dispersion of the acoustic waves (Nussenzveig 1972, O’Donell et al 1978) contribute to
the broadening of the optoacoustic signals (Patch and Greenleaf 2007, Treeby and Cox
2009), which in turn causes loss of resolution of the reconstructed images and quantification
inaccuracy. The resolution of the images depends on the available frequency content of
the ultrasonic signals used to make the reconstruction, which is also determined by other
factors apart from the acoustic attenuation. For example, the frequency response (FR) of
the transducer or filtering of the measured signals may also affect the broadening and their
contributions could become stronger than the effect of acoustic attenuation.
Due to its frequency-dependence, acoustic attenuation is usually treated in the frequency
domain. In ultrasonic imaging, the frequency bandwidth of the signals is relatively narrow
and usually only the reduction of amplitude is considered as distortion effect, so that the
correction is made with the so-called time-gain compensation (Wells 1999). However, in
optoacoustic imaging, the ultrasonic waves are generated by thermal expansion of structures
with varying dimensions resulting in signal bandwidth that may be very broad, for example
ranging from 50 Hz to 50 MHz (Gusev and Karabutov 1993). In this case it would be
important to consider the effects of acoustic attenuation in the detected optoacoustic signals.
A lossy operator can be added to the time domain wave equation so that both the frequency
dependence of the attenuation coefficient and the causality imposed conditions are fulfilled
(Treeby and Cox 2010). Based on this modified equation, an algorithm for correcting the
effects of wave propagation in a uniform attenuating medium can be derived (La Riviere
et al 2006, Burgholzer et al 2010). Other methods to avoid the acoustic attenuation effects
consist in calculating the ratio of spectra of the signals corresponding to two different laser
wavelengths (Guo et al 2010) or in a modification of the time-reversal reconstruction algorithm
(Treeby et al 2010). The space-dependent attenuation of the sample can be estimated with
a system integrated in the optoacoustic tomographic setup (Jose et al 2011). Since in many
optoacoustic setups the samples may be immersed in a water tank, it may be important to
consider the effects of acoustic attenuation in both tissue and in water to make an appropriate
correction.
The goal of this work is three-fold. First, we examine the effects of acoustic attenuation
on the optoacoustic signals to identify the areas where optical and acoustic corrections are
practically necessary for improving the derived optoacoustic images. Second, we investigate
the relative merits of acoustic attenuation versus other factors, such as optical attenuation,
FR of the transducer and non-uniform speed of sound, to establish in which cases the
effects of acoustic attenuation are dominant. Finally, we propose a correction algorithm
The effects of acoustic attenuation in optoacoustic signals 6131

for accounting for the effects of ultrasonic attenuation and dispersion, applicable in both the
time and frequency domains, and demonstrate imaging benefits in experimental measurements.

2. The effects of acoustic attenuation

We review in this section the theoretical effects of acoustic attenuation in ultrasonic


waves propagating in tissue and in water by using a model consistent with experimental
measurements. Based on these theoretical results, we will then compare the effects of acoustic
attenuation with the effects of other factors that need to be considered in practical situations.

2.1. Modelling of acoustic attenuation


The pressure caused by a monochromatic acoustic plane wave of angular frequency ω in a
uniform attenuating medium at a point x and at an instant t can be expressed as
p(x, t) = Re[p̂(x, t)] = Re{p̂0 exp[j (ωt − k̂x)]}, (1)
where p̂0 is the complex amplitude and k̂ is the complex wavenumber defined as
ω
k̂ = − j α(ω), (2)
c(ω)
where c(ω) is the phase velocity and α(ω) the attenuation coefficient. It has been established
empirically that the attenuation coefficient in a wide variety of lossy media satisfies a power
law dependence on frequency of the form (Sushilov and Cobbold 2004)
α(ω) = α0ω |ω|n = α0f |f |n , (3)
where n is a real positive constant.
In a medium with an attenuation coefficient satisfying equation (3), the acoustic waves
must be dispersive in order to guarantee causality, i.e. to guarantee that the cause precedes
the effect (Nussenzveig 1972). Then, the associated dependence of the phase velocity on
frequency is determined by the Kramers–Kronig relationships and can be expressed as a
function of the phase velocity c(ω0 ) for a reference angular frequency ω0 as (Waters et al
2005)
⎧ πn
⎪ 1

⎨ + α0ω tan (|ω|n−1 − |ω0 |n−1 ) for 0 < n < 1 or 1 < n < 3
1 c(ω0 ) 2
=   (4)
c(ω) ⎪ ⎪ 1 2 ω
⎩ − α0ω ln   for n = 1
c(ω0 ) π ω0

2.2. Reduction of amplitude


The primary effect of acoustic attenuation is the reduction of the amplitude of the waves. The
amplitude of a monochromatic acoustic plane wave is calculated as pm (x) = mod[p̂(x, t)],
where p̂(x, t) is given by equation (1). Then,
pm (x) = pm (0) exp[−αT (ω, x)], (5)
being αT (ω, x) = α(ω)x the total attenuation. Experimental measurements in water indicate
that the value of n in equation (3) is very close to 2 over a wide range of frequencies
and the value of the attenuation constant α0f = 0.00217 dBMHz−2 cm−1 (Szabo 2004).
On the other hand, for many biological tissues the value of n is approximately equal to 1 and
the standard value of the attenuation constant α0f ≈ 0.5 dBMHz−1 cm−1 for the frequency
range of ultrasonic imaging (Szabo 2004). For comparison purposes, figure 1(a) displays an
6132 X L Deán-Ben et al

(a) (b) (c)

Figure 1. The effects of acoustic attenuation for standard tissue (continuous lines) and for water
(dashed lines). (a) Attenuation coefficient as a function of frequency. (b) Phase velocity as a
function of frequency. (c) Broadening of the optoacoustic signals as a function of the propagating
distance.

average attenuation coefficient for tissue and for water as a function of frequency estimated
with equation (3). It is shown that the attenuation coefficient is much higher in tissue than in
water for the entire frequency range. As observed on figure 1(a), the water contribution can
be practically neglected for frequencies up to 20–30 MHz.

2.3. Broadening
For broadband ultrasonic waves, since each frequency component is attenuated differently in
an attenuating medium, the frequency spectrum detected after wave propagation through the
attenuating tissue narrows, and consequently the width of the time-domain signals broadens.
The dispersion of the acoustic waves, i.e. the dependence of the phase velocity on frequency is
given by equation (4) for average tissues (n = 1) and for water (n = 2). Figure 1(b) displays
the phase velocity as a function of frequency for a standard tissue and for water calculated
considering a standard value of the phase velocity c(f0 ) = 1500 m s−1 for a frequency
f0 = 1 MHz.
The broadening of the signals due to acoustic attenuation can be estimated by assuming
a plane wave for which the initial pressure is given by a Dirac’s delta, i.e. p̂0 = 1 (in arbitrary
units) in equation (1) for all the frequencies of the spectrum. Then, recalling the principle of
superposition, valid for linear acoustics, the wave is equivalent to the superposition of all its
frequency components (equation (1)), i.e. the pressure at a point x and at an instant t can be
expressed as

1 ω
p(x, t) = Re exp[−α(ω)x] exp −j x exp(j ωt) dω . (6)
2π −∞ c(ω)
Then, the frequency dependence of the attenuation coefficient (equation (3)) and of the
phase velocity (equation (4)) must be substituted in equation (6).
For standard tissue (n = 1 in equation (4)), the dispersion of the acoustic waves is very
low and a constant phase velocity ct is usually considered. In such a case, an analytical solution
of equation (6) can be derived, which is given by La Riviere et al (2005)
α0ω,t x
p(x, t) = , (7)
π [(α0ω,t x)2 + (x/ct − t)2 ]
where α0ω,t is the attenuation constant in tissue. Numerical calculations of equation (6)
considering a constant phase velocity and a phase velocity given by equation (4) show that the
broadening of the acoustic signals is similar in both cases (Szabo 2004). Then, we estimate
The effects of acoustic attenuation in optoacoustic signals 6133

the width of the acoustic signals as a function of the propagating distance in tissues tt as the
full width at half maximum (FWHM) of equation (7), i.e.
tt = 2α0ω,t x. (8)
For water (n = 2 in equation (4)), the phase velocity cw is constant. Then, an analytical
solution of equation (6) can also be derived, which is given by Patch and Greenleaf (2007)

1 π −(x/cw − t)2
p(x, t) = exp , (9)
2π α0ω,w x 4α0ω,w x
where α0ω,w is the attenuation constant in water. The width of the acoustic signals as a function
of the propagating distance in water tw is estimated as the FWHM of the exponential term
in equation (9), i.e.

tw = 4 α0ω,w x ln 2. (10)
The broadening of the acoustic signals as a function of the propagating distance in water
and in tissue for distances typical in optoacoustic imaging, i.e. 0–5 cm, is shown in figure 1(c),
estimated with equations (8) and (10), respectively. The FR of a focused immersion transducer
is conditioned by the acoustic attenuation of the waves in water through a distance equal to
the focal distance of the transducer, so it is the attenuation in tissue what mainly increases the
width of the signals. Then, we can neglect the effects of water in the broadening of the signals
in many practical cases.
The width of the acoustic waves generated by illuminating an optical absorber with a
short-pulsed laser is proportional to the size of the absorber as t = x/c (Wang 2009),
so that the resolution of the images obtained by assuming a homogeneous, non-attenuating
medium is limited to x when the width of the signals is increased t.

3. Acoustic attenuation in practical situations

In this section, we derive a comparison between the effects of acoustic attenuation and the
effects of other factors such as the optical attenuation, the FR of the transducer and the speed of
sound variations, which also need to be taken into account in actual measurements. Thereby,
we determine in which cases acoustic attenuation is the dominant cause of distortion in the
optoacoustic signals so that it is convenient to make a correction for its effects. The correction
procedure will be introduced in the next section.

3.1. Acoustic attenuation versus optical attenuation


In quantitative optoacoustic imaging, the ultimate goal is to estimate the distribution of the
optical absorption coefficient, which relates to the concentration of chromophores. The
reconstructed magnitude is however the initial optoacoustic pressure, which is also affected
by the light fluence distribution. Thereby, the non-uniform light fluence distribution caused by
optical attenuation must be taken into account in order to estimate the distribution of the optical
absorption coefficient. The reduction of amplitude of the waves due to acoustic attenuation
also leads to errors in the estimated distribution of the optical absorption coefficient, so that this
effect may also play a role for quantification purposes. We derive in this section a comparison
of the reduction of amplitude caused, respectively, by acoustical and optical attenuation as a
function of the frequency of the acoustic waves.
The distribution of light fluence U ( x ) inside a tissue under continuous wave (CW)
illumination satisfies the diffusion equation given by Razansky and Ntziachristos (2007)
−1
  ∇ 2 U ( x ) = q0 (
x ) + μa U ( x ), (11)
3 μs + μa
6134 X L Deán-Ben et al

being μa the absorption coefficient and μs the reduced scattering coefficient. Typical values
of tissue are μa = 0.2 cm−1 and μs = 10 cm−1 (Razansky and Ntziachristos 2007). Although
equation (11) is derived for CW light, it has been successfully applied in optoacoustic imaging
with pulsed-light illumination (Jetzfellner et al 2009).
For a semi-infinite medium with no distributed light sources and with uniform illumination
on the surface, a one-dimensional analytical solution of equation (11) [U (x)] can be readily
calculated by considering that the source term q0 (x0 ) = 0 and by modelling the illumination
by means of the boundary conditions U (0) = U0 and U (∞) = 0. In such a case, it is verified
that
U (x) = U (0) exp(−μt x), (12)


where μt = 3μa (μs + μa ).
The initial optoacoustic pressure poa (x ) calculated with an optoacoustic system relates to
the light fluence established in the tissue imaged by Köstli et al (2001)
x ) = μa (
poa ( x )U (
x ), (13)
where  is the dimensionless Grüneisen parameter. In a semi-infinite medium with constant
illumination on the surface and with constant μa , the optoacoustic pressure can then be
obtained by combining equations (12) and (13), i.e.
poa (x) = poa (0) exp(−μt x). (14)
The amplitude of an acoustic wave generated at a depth x from the surface of a semi-
infinite medium is further reduced as the wave propagates to the surface due to the acoustical
attenuation. For average tissue, the amplitude of the acoustic wave is obtained as a function
of the propagation distance by combining equations (3) and (5), i.e.
pm (x) = pm (0) exp(−α0f f x). (15)
By comparing equations (14) and (15) we can estimate the frequency fc for which the
effect of acoustic attenuation in the reduction of amplitude of the optoacoustic signals is the
same as the effect of optical attenuation. It is given by
μt
fc = . (16)
α0f
Therefore, the relative optical and acoustic attenuation contributions depend on the
frequency bandwidth that is captured by the optoacoustic system utilized. If the frequency
bandwidth of the captured signals is around fc , both optical and acoustical attenuation
contribute similarly to the error in the estimated distribution of the optical absorption
coefficient. Considering the standard values μa = 0.2 cm−1 , μs = 10 cm−1 and
α0f = 0.5 dBMHz−1 cm−1 , the frequency is calculated to be fc ≈ 43 MHz.

3.2. Acoustic attenuation versus frequency response of the transducer


The broadening of the optoacoustic signals directly relates to a loss of resolution in the
reconstructed images. We have seen in section 2.3 that such broadening can be caused by
acoustic attenuation. However, any other phenomenon that narrows the frequency spectrum
of the captured signals may also contribute to the increment in the width of the signals. In
particular, the FR of the acoustic transducer must also be considered in practical situations.
We derive in this section a comparison of the broadening of the signals caused, respectively,
by acoustical attenuation and by the FR of the transducer.
The amplitude of the FR of the transducer corresponds to its sensitivity to the detection
of the spectral components of a given signal. In common piezoelectric transducers, the FR is
The effects of acoustic attenuation in optoacoustic signals 6135

mainly determined by a mechanical resonance of the piezoelectric cristal. Thereby, the shape
of the FR corresponds to a smooth curve (typically similar to a Gaussian curve) with a peak at
the frequency fct corresponding to the maximum sensitivity of the transducer. The bandwidth
of the transducer ft is usually defined from the upper and lower −6 dB frequencies with
respect to fct . For a given shape of the FR, it is verified that
ft = k  fct , (17)
being k  a constant. As mentioned above, the FR of the transducer also contributes to the
broadening of the signals. The impulse response (IR) of the transducer is given by its response
to an impulse-type wave in the time domain. For a transducer with a central frequency fct , the
IR corresponds to a wavetrain with a central frequency fct modulated by an envelope whose
width tFR is inversely proportional to the bandwidth of the transducer, i.e.
k  k
tFR = 
= . (18)
k fct fct
The width of the envelope in the time domain for a measured impulse-type wave
corresponds to the broadening due to the FR of the transducer. The constant k in
equation (18) depends on the shape of the FR of the transducer. For a Gaussian shape,
k  ≈ 1, and for common piezoelectric transducers ft ≈ fct , so that k ≈ 1. By comparing
equations (8) and (18), the broadening of the acoustic signals due to the FR of the transducer
and due to the acoustic attenuation in tissues is the same when
α0f xfct = π k. (19)
The first term in equation (19) corresponds to the total acoustic attenuation of the waves
for the central frequency of the transducer. Then, the effect of the FR of the transducer is
dominant in the broadening of the acoustic signals when the acoustic attenuation for fct is
lower than π k, where k depends on the transducer. The FR of the transducer affects equally
to waves generated at different points whereas the acoustic attenuation for fct depends on the
imaging depth. Equivalently, equation (19) indicates that for a given depth x the broadening
of the signals due to acoustic attenuation is the same as the broadening of the signals due to
the FR of the transducers. Then, the loss of resolution of the images is mainly determined by
acoustic attenuation for points located deep into the tissue and by the FR of the transducer for
points located close to the surface.

3.3. Acoustic attenuation versus space-dependent speed of sound


When signals are acquired over multiple angles (projections) around the sample imaged, as
typical in some tomographic systems (Ma et al 2009), the resolution is also conditioned by
the difference in the speed of sound corresponding to different regions within tissue. In
this section, we establish a comparison of the loss of resolution caused by speed of sound
heterogeneities and by acoustic attenuation.
A space-dependent speed of sound causes time-shifting of the signals with respect to the
signals corresponding to ultrasonic waves propagating solely in water (Xu and Wang 2003).
For example, when the signals propagate a distance lt in a tissue the time-shift tc of the
signals is given by
 
1 1
tc = − lt , (20)
ct cw
where cw is the speed of sound in water and ct is the speed of sound in tissue.
Equation (20) refers to the case in which the waves propagate along straight acoustic rays
6136 X L Deán-Ben et al

Figure 2. Optoacoustic setup for acquisition along a circumference with radius R (dashed line),
in which a point optical absorber is located at the centre. An area with radius Rt having a different
speed of sound is present (area inside the dotted circumference). The arc centred at the position of
the transducer (dashed line) indicates the contribution of the signal acquired at the transducer to the
reconstruction when assuming constant speed of sound, so that the final result of the reconstruction
in such a case is the circumference (continuous line) indicated in the figure.

(we neglect refraction), which is a valid approximation to estimate the time-shift due to small
variations in the speed of sound, as is the case in soft biological tissues (Deán-Ben et al 2011b).
In the case that the speed of sound within the tissue is not uniform, 1/ct is the mean value of
the inverse of the speed of sound.
If no correction is made to the reconstruction algorithm, the images are distorted and so
the resolution is reduced. As a representative example to understand the effect of a region with
different speed of sound in terms of loss of resolution, we consider that the signal is acquired
along a circumference with radius R and that there is a point source located at its centre
(figure 2). We assume that the set-up is immersed in water, with speed of sound cw , and that a
circular region with radius Rt concentric with the acquisition circumference and with a speed
of sound ct does exist. In such a case, the optoacoustic signal is detected at each transducer at
instant tt = (R − Rt )/cw + Rt /ct . If a uniform speed of sound equal to cw is considered for the
reconstruction, the source of the wave is assumed to be located at a distance cw tt away from the
transducer (figure 2). Then, the reconstructed image obtained by adding the contributions of
all the positions of the transducer consists of a circumference with diameter equal to cw |tc |
(figure 2), where tc is given by equation (20) by considering lt = 2Rt . We can consider as an
approximation in a general case that the reconstructed image corresponding to a point source
consists of a geometrical shape with size x = cw |tc |, being tc given by equation (20) by
considering lt as the characteristic length of the area with different speed of sound. The size
of the geometrical shape can be considered as a limitation for the achievable resolution in the
images. A more detailed analysis of the effects of speed of sound heterogeneities can be found
in Xu and Wang (2003).
Therefore, by comparing equations (8) and (20) we deduce that the effects of acoustic
attenuation and the effects of speed of sound variations are equivalent if
 −1
1
ct = ± 2α0ω . (21)
cw
The effects of acoustic attenuation in optoacoustic signals 6137

For a standard speed of sound in water cw ≈ 1500 m s−1 and a standard attenuation
constant in tissue α0f = 0.5 dBMHz−1 cm−1 , the value of ct in equation (21) is 1496 m s−1
or 1504 m s−1 , which corresponds to a speed of sound variation of approximately 0.27%. In
practice, the variations in the speed of sound in different tissues are usually in the range ±10%
(Szabo 2004), so that in principle the loss of resolution in tomographic setups is mainly due
to speed of sound variations in most practical cases. However, it is important to take into
account that the loss of resolution due to speed of sound variations is estimated for the case
in which a uniform speed of sound equal to cw is considered. The resolution may improve if
the reconstruction is done with a speed of sound representing the variations in the propagating
region. This issue will be discussed in a future work.

4. Attenuation correction

We establish in this section a relationship between the optoacoustic pressure caused by the
propagation of the wave in a uniform attenuating medium and the optoacoustic pressure
caused by its propagation in a non-attenuating medium. The derivation is performed from the
wave equation in the frequency domain, yielding an equivalent result as in Szabo (2004) and
La Riviere et al (2006). Then, based on the resulting expression for a uniform attenuating
medium, we suggest a correction for the signals corresponding to the propagation of the
waves in non-uniform attenuating media, i.e. in media with different regions having different
attenuation behaviours. The correction corresponding to non-uniform attenuating media
usually must be considered in practice, for example in samples immersed in water.
Taking into account the complex wavenumber k̂ defined in equation (2), the wave equation
in the frequency domain (Helmholtz equation) for a uniform attenuating medium with a point
source located at the origin can be expressed as
∇ 2 p̃( x , ω) = δ(
x , ω) + k̂ 2 p̃( x ). (22)
The solution of equation (22) in a 3D domain is given by
−1
x , ω) =
p̃( exp(−j k̂|x |)
4π |x|
 
−1 ω
= exp −j | x | exp[j γ (ω)|
x |]
4π |x| c0
= p̃0 (
x , ω) exp[j γ (ω)| x |], (23)
being
1 1
γ (ω) = ω − − j α(ω). (24)
c0 c(ω)
The term p̃0 (
x , ω) corresponds to the equivalent solution for a non-attenuating medium.
Then, the effects of acoustic attenuation for each component of the spectrum of the signal can
be corrected with equation (23), taking into account the distance between the source and the
measuring point. An equivalent expression in the time-domain can be obtained taking into
account the principle of superposition. Thereby, p( x , t) = Re[p̂(
x , t)], where p̂(
x , t) is given
by

1
x , t) =
p̂( p̃(
x , ω) exp(j ωt) dω
2π −∞

1
= p̃0 ( x |] exp(j ωt) dω.
x , ω) exp[j γ (ω)| (25)
2π −∞
6138 X L Deán-Ben et al

Equivalently, equation (25) can be expressed as


x , t) = p̂0 (
p̂( x , t) ∗ aifr(
x , t), (26)
where ∗ denotes convolution and the term aifr(
x , t) stands for the attenuation impulse FR,
which is given by
x , t) = −1 {exp[j γ (ω)|
aifr( x |]}, (27)
−1
where  is the inverse Fourier transform (FT). Therefore, the attenuation correction can also
be regarded as a deconvolution in the time domain.
In practical situations, the attenuation is space-dependent. Even when a uniform
attenuating sample is imaged, the signals are usually collected with immersion transducers,
and water presents different attenuation behaviour as discussed above. If there is no significant
acoustic mismatch between the different regions through which the acoustic waves propagate,
it is reasonable to model the propagation of the waves considering straight acoustic rays and
we assume that equation (23) can be modified as follows

x , ω) = p̃0 (
p̃( x , ω) exp j γ (ω) dl , (28)
l

where l is a straight line between the acoustic source and the point in which the pressure
is measured. Taking into account equation (28), the term aifr(x , t) in equation (26) can be
expressed as

x , t) = −1 exp j
aifr( γ (ω) dl , (29)
l

5. Materials and methods

To experimentally verify the theoretical effects of acoustic attenuation and to evaluate the
performance of the correction we made several experiments.
As the first step, we compared the measured amplitude and width of the optoacoustic
signals generated at a point source when the waves propagate solely in water and when a
phantom of attenuating tissue is introduced between the source and the transducer, so that the
effects of acoustic attenuation in terms of reduction of amplitude and broadening of the waves
can be evaluated. In such a case, neglecting the attenuation in water, acoustic attenuation
takes place exclusively in the phantom. In order to estimate the frequency dependence of the
attenuation, we compared the Fourier spectra of the signals captured with and without acoustic
attenuation. Specifically, a linear dependence of the attenuation coefficient with frequency
is expected in tissues, so that the attenuation constant α0f can be estimated by fitting the
frequency dependence of the attenuation coefficient to a straight line.
The feasibility of the correction procedure was tested with the signals collected in the
presence of the attenuating phantom. The correction is done in the frequency domain. First,
we performed the correction of the amplitude of the spectral components by assuming an
attenuation coefficient linearly dependent on frequency. The phase of the signals in this case
it is not corrected. Then, we estimated the attenuation coefficient and the parameter n in
equation (3) by fitting the dependence of the attenuation coefficient on frequency to a power
law. In this second case we also considered the undistorted phase (i.e. the phase of the signal
measured without the attenuation phantom in the set-up). By comparing the performance
of both corrections in terms of amplitude and width of the signals we can conclude on the
feasibility of performing the correction by assuming a linear dependence of the attenuation
coefficient and not correcting the phase distortions. This is an important conclusion as in
The effects of acoustic attenuation in optoacoustic signals 6139

practice the parameter n may not be readily determined and may vary between different
regions. Also, the possible correction of the phase is rather tricky in practical situations. First,
equation (4) is given as a function of the phase velocity at a reference frequency, which must
be established heuristically. Moreover, the effect of the difference in the speed of sound in
tissue with respect to water also contributes to a phase variation and must be corrected as well.
As the second step, we performed another experiment to analyse how the reduction of
amplitude and broadening of the signals affect the optoacoustic tomographic reconstructions
as well as to assess the feasibility of a correction. For this, we imaged a phantom with
embedded microparticles in two possible scenarios, namely with and without an attenuating
phantom between the imaging phantom and the transducer. This corresponds to a simple case
in which optical attenuation does not affect the reconstruction (the phantom is transparent)
and the wave generated at any point of the phantom propagates through the same distance
of attenuating tissue. However, this experiment allows assessing the improvement of the
reconstructed images when the correction of the attenuation effects is made as it is possible to
compare the corrected image with the image obtained with no attenuation. The tomographic
reconstructions were obtained with the delay-and-sum algorithm (Hoelen and de Mul
2000).

5.1. Experimental setup


The experimental set-up to measure the optoacoustic signals generated at a point source
is depicted in figure 3. The output light beam was directed to a poly(methyl methacrylate)
(PMMA) slab with one face painted in black, and the generated ultrasonic waves were detected
with a piezoelectric transducer. Both the slab and the transducer were immersed in a water
tank. The laser emits 10 ns pulses with a wavelength equal to 605 nm and with 15 MHz pulse
repetition rate. We assume that the absorption of light takes place uniformly throughout the
layer of paint, i.e. the optoacoustic source was approximately a cylinder with a diameter equal
to the diameter of the laser beam (approximately 500 μm in our case) and a height equal to
the thickness of the layer of paint (approximately 50 μm in our case). Such an optoacoustic
source can be considered a point source for the frequency range of the transducers that we
employed in the experiment. The width of the PMMA slab was large enough to prevent
the reflected acoustic waves at its opposite face, interfering with the measured signals. Two
cylindrically focused immersion transducers were used to collect the ultrasonic signals, which
were amplified and digitized by an embedded oscilloscope card at a sampling rate of 100 MSPS
and 14 bit resolution. The signals were averaged 64 times in order to improve the signal-
to-noise ratio. The central frequencies of the ultrasonic transducers were 3.5 and 7.5 MHz
and their focal lengths were, respectively, 38.1 and 25.4 mm. As indicated in figure 3, an
attenuating phantom can be introduced in between the slab and the ultrasonic transducer, so
that we collected the ultrasonic signals with and without the phantom in the setup.
The agar phantom with embedded microparticles was imaged with the experimental
system described in detail in Ma et al (2009). Basically, the output beam of the aforementioned
laser was split into two parts and directed onto the object from two opposite sides to attain a ring-
type uniform illumination conditions on the surface of the object. The generated ultrasonic
waves were measured with the aforementioned acquisition system in which the 7.5 MHz
transducer was used. The sample was supported in a rotation stage, and the ultrasonic signals
corresponding to 180 tomographic projections were collected by rotating the sample with
angular steps of 2◦ . For each projection, the signal was averaged 64 times. A band-pass filter
with lower and upper cut-off frequencies 0.25 and 15 MHz, respectively, was applied to the
acquired signals.
6140 X L Deán-Ben et al

Figure 3. Experimental setup for acquiring the point optoacoustic signals described in section 5.1.

5.2. Phantoms
In the experiment corresponding to the measurement of the optoacoustic signals generated in
a point source we employed as attenuating phantoms four pork fat samples with respective
thicknesses 6, 11, 17 and 29 mm.
On the other hand, the imaging phantom consisted of a 3 mm cylinder of pure agar
solution (1.3% agar powder by weight). Polyethylene microparticles with an approximate
diameter of 50 μm (Cospheric BKPMS 45–53 μm) were inserted in the imaged slice of the
phantom as optical absorbers. The attenuating phantom consisted again of pork fat sample, in
this case with thickness 7 mm.

6. Results

6.1. Optoacoustic signals


The signals obtained with the point optoacoustic source described in section 5.1 are shown
in figures 4(a) and (b) for the transducers with central frequencies of 3.5 and 7.5 MHz,
respectively. The signals were obtained with no attenuating phantom in between the source
and the transducer (black lines) and by introducing the fat samples with respective thicknesses
6 mm (blue lines), 11 mm (red lines), 17 mm (green lines) and 29 mm (yellow line). The
reduction of amplitude of the waves when introducing the attenuating phantoms is clearly
perceived in the signals. The amplitudes of the optoacoustic signals were estimated as the
maximum in the modulus of their Hilbert transform [H (p)(t)]. The values are displayed
in table 1 (amplitude of the measured signals). On the other hand, the broadening of the
signals when introducing the attenuating phantoms can be perceived in the moduli of the
Hilbert transforms (normalized) shown in figures 4(c) and (d). The widths of the signals were
estimated as the FWHM of the moduli of the Hilbert transforms. The values are also displayed
in table 1 (width of the measured signals). For small attenuation, the broadening of the signals
is relatively low, which indicates that the widths of the signals are mainly determined by the
FR of the transducer. However, for larger values of the attenuation, the widths of the signals
clearly increase, which corresponds to the broadening due to acoustic attenuation.
Figures 4(e) and (f) show the moduli of the FTs of the signals. One can clearly see that
the high frequency components of the signals are more attenuated according to the frequency-
dependence of the acoustic attenuation. The total attenuation αT (f ) (equation (5)) given in
dB as a function of frequency is estimated as
 
p̃0 (f )
αT (f ) = 20 log . (30)
p̃(f )
The effects of acoustic attenuation in optoacoustic signals 6141

(a) (b)

(c) (d)

(e) (f)

(g) (h)

Figure 4. (a) and (b) Signals acquired with the transducers with central frequencies of 3.5 and
7.5 MHz, respectively. (c) and (d) Modulus of the Hilbert transforms (normalized) of (a) and
(b), respectively. (e) and (f) Modulus of the Fourier transforms of (a) and (b), respectively. The
signals are acquired with no attenuating tissue in the setup (black lines), with the fat sample 6 mm
thick (blue lines), with the fat sample 11 mm thick (red lines), with the fat sample 17 mm thick
(green lines) and with the fat sample 29 mm thick (yellow lines). (g) and (h) Total attenuation as
a function of frequency. The colour code correspond to the signals above.
6142 X L Deán-Ben et al

Table 1. Quantification of the measured optoacoustic signals (with and without acoustic
attenuation) and the corresponding signals being corrected for the effects of acoustic attenuation
considering an attenuation coefficient linearly dependent on frequency. α0f is the attenuation
constant estimated by assuming a linear dependence on frequency and flp is the cut-off frequency
of the low pass filter employed to make the correction.

Measured signals Corrected signals


Fat sample Transducer Amplitude Width α0f flp Amplitude Width
(mm) (MHz) (A.U.) (ns) (dBMHz−1cm−1) (MHz) (A.U.) (ns)

– 3.5 0.1104 238 – – – –


6 3.5 0.0607 267 2.975 7.5 0.1096 251
11 3.5 0.0334 319 3.283 7.5 0.1037 281
17 3.5 0.0199 369 3.404 7.5 0.1034 287
29 3.5 0.0095 472 3.295 5 0.0819 358
– 7.5 0.1210 126 – – – –
6 7.5 0.0409 164 3.457 11.5 0.1203 138
11 7.5 0.0261 218 3.174 8.5 0.1064 172
17 7.5 0.0146 293 3.029 7.5 0.0708 204

The values of αT (f ) are represented in figures 4(g) and (h) for the points in which
the amplitude of the Fourier spectrum of the attenuated signals is higher than 1/10 of its
maximum value. The attenuation constant α0f (equation (3)) was first estimated by assuming
an attenuation coefficient linearly dependent on frequency. For this, the points in figures 4(g)
and (h) were fitted to straight lines having a zero ordinate at the origin (continuous lines in
these figures). Specifically, the attenuation coefficient is calculated as α0f = m/ lt , being m
the slope of such lines (in dB/MHz) and lt the thickness of the fat sample. The values of α0f
estimated with this approach for the attenuating phantoms employed in the experiment are
displayed in table 1 (α0f of the measured signals). The attenuation constant in our samples is
much higher than the experimentally measured attenuation constant in standard tissue.
The correction of the attenuation effects for the signals in figures 4(a) and (b) was made by
considering a constant γ (ω) in equation (28). The term γ (ω) has two components, −j α(ω)
corresponds to the correction of the reduction of amplitude of the spectral components of a
given signal, whereas ω[1/c0 − 1/c(ω)] represents the correction of the dispersion associated
with attenuation, which causes a change in the phase of the Fourier spectrum.
In a first step, we considered an attenuation coefficient linearly dependent on frequency
and we neglected the dispersion effects, i.e. we did not correct the phase of the signals.
Figures 5(a) and (b) show two examples of the measured signals with attenuation (blue lines)
and without attenuation (black lines). The moduli of their FTs correspond, respectively, to
the blue and black lines in figures 5(c) and (d). The corrected spectra of the signals (green
lines in figures 5(c) and (d)) were obtained with equation (28). One unwanted effect of the
correction (especially noticeable in figure 5(d)) is that the amplification of the high frequency
components causes the amplification of the noise. To avoid this effect, a low pass filter
with cut-off frequency flp is applied, resulting in the red dashed lines in figures 5(c) and
(d). The values of the cut-off frequencies flp used for the correction of each signal are
displayed in table 1. Then, the corrected signals are obtained as the real part of the inverse
FTs of the red dashed lines and are shown in figures 5(a) and (b) (red lines). The estimated
amplitudes and widths of the corrected signals are obtained by means of the Hilbert transform
The effects of acoustic attenuation in optoacoustic signals 6143

(a) (b)

(c) (d)

Figure 5. (a) and (b) Signals acquired and corrected corresponding, respectively, to the transducers
with central frequencies 3.5 and 7.5 MHz. The signals are acquired without fat sample (black
lines) and with fat sample (blue lines). The fat sample is 6 mm thick in (a) and 17 mm thick in
(b). The red lines and the green lines correspond to the corrected signals obtained, respectively,
by considering an attenuation coefficient linearly dependent on frequency and no distortion effects
and by considering an attenuation coefficient satisfying a power law dependence on frequency and
the original phase of the signals. (c) and (d) Acquired and corrected modulus of the Fourier spectra.
The black and blue lines correspond to the modulus of the Fourier transform of the black and blue
lines in (a) and (b). The green lines correspond to the corrected modulus of the Fourier spectra
calculated considering equation (28) by assuming a linearly dependent attenuation coefficient. The
dashed red lines correspond to the low pass filtered green lines. The correction is made considering
the real part of the inverse Fourier transform of the dashed red lines.

as described above. The values are displayed in table 1 (amplitude and width of the corrected
signals).
To analyse the effect of a more accurate correction we considered in a second step that
the attenuation coefficient satisfies a generic power law dependence on frequency and that
the phase of the corrected signals corresponds to the original phase (phase of the FT of the
signal with no attenuation). The values of n (equation (3)) and the attenuation constant α0f (in
units dBMHz−n ) are obtained by fitting the points in figures 4(g) and (h) to a curve satisfying
a power law dependence (dashed lines in these figures). They are shown in table 2. The
corrected signals are obtained with the same filters in the Fourier spectrum as in the previous
case (flp in table 1). The corrected signals obtained with this approach for the example in
figure 5 correspond to the green lines in figures 5(a) and (b). The amplitudes and widths
of the corrected signals are displayed in table 2. They are also estimated by means of the
6144 X L Deán-Ben et al

Table 2. Quantification of the corrected optoacoustic signals by considering an attenuation


coefficient satisfying a generic power law dependence on frequency and the original phase of
the signals measured without attenuation.

Fat sample Transducer α0f Amplitude Width


(mm) (MHz) n (dBMHz−n) (A.U.) (ns)

6 3.5 1.185 1.366 0.1080 244


11 3.5 1.202 2.785 0.1058 257
17 3.5 1.068 5.317 0.1116 249
29 3.5 1.101 8.695 0.952 281
6 7.5 1.191 1.447 0.1213 136
11 7.5 1.318 2.078 0.1082 167
17 7.5 1.310 3.432 0.0860 178

Hilbert transform as in the previous case. The values of the amplitude and the width of the
corrected signals obtained with this approach are displayed in table 2. Although there is an
improvement in terms of amplitude and width of the signals (they are closer to the amplitudes
and widths of the measured signals without attenuation), in general the noise in the corrected
signals is higher than the noise in the corrected signals obtained with the procedure described
above. This is due to the higher amplification of the high frequency components of the noise.
To reduce the noise, we need to use a filter with a lower cut-off frequency, in which case the
correction of the amplitude and the width would be worse. In any case, the improvement with
respect to the first order approximation described above is not very significant. Therefore, in
many practical situations it may be convenient to use a model based on a linear dependence of
the attenuation coefficient and no distortion of the phase, which is much easier to implement.

6.2. Image reconstruction


The tomographic reconstruction of the phantom described in section 5.2 obtained with the
delay-and-sum algorithm is shown in figure 6(a). When introducing the pork fat sample in
the setup, the signals not only are affected by acoustic attenuation but are also time-shifted
(equation (20)) due to the acoustic mismatch. Thereby, the signals collected when the pork
fat sample is introduced in the setup are time-shifted in order to correct for the acoustic
mismatch. Then, the corresponding tomographic reconstruction obtained with the delay-
and-sum algorithm is shown in figure 6(b). The signals are then corrected for attenuation
effects as described in section 4. An attenuation coefficient linearly dependent on frequency
is assumed and no dispersion effects are considered. A low pass filter with cut-off frequency
flp = 9.5 MHz is applied to the corrected signals. The resulting tomographic reconstruction
obtained with the corrected signals is shown in figure 6(c).
The loss of resolution in figure 6(b) with respect to figure 6(a) is clearly perceived in
the images. Such loss of resolution is partially corrected in figure 6(c) at the expense of
introducing background noise due to the amplification of the relatively high frequency noise.
A better insight into the attenuation effects is gained by comparing the profiles A and B marked
in figure 6(a), which are shown in figures 6(d) and (e), respectively. Apart from the loss of
resolution, in such profiles the effect of the reduction of amplitude in the image is perceived,
which is also partially corrected with the attenuation correction. Figures 6(f) and (g) show the
normalized profiles A and B, respectively, which allow assessing the resolution changes. In
profile A we can clearly distinguish three peaks when there is no acoustic attenuation whereas
The effects of acoustic attenuation in optoacoustic signals 6145

(a) (b) (c)

1 mm 1 mm 1 mm

(d) (e) (f) (g)

Figure 6. Tomographic reconstruction of the agar phantom described in section 5.2. (b) Equivalent
reconstruction obtained when the signals are acquired in the presence of a fat sample between the
phantom and the transducer. (c) Tomographic reconstruction obtained with the same signals as in
(b) after correcting them for the effects of acoustic attenuation. (d) and (e) Profiles A and B shown
in (a) for (a) (continuous lines), (b) (dashed lines) and (c) (dotted lines). (f) and (g) Equivalent
normalized profiles.

only two peaks are detected when introducing the attenuating tissue. In the corrected profile,
the three peaks can be distinguished and the peak in the left pops up. In profile B, two peaks
are distinguished with no attenuation and only one with attenuation. Although the second
peak is not distinguished in the corrected image, the width of the larger peak is reduced, which
indicates an increase in the resolution.

7. Discussion and conclusions

In this paper we reviewed the theoretical effects of acoustic attenuation, namely the reduction
amplitude and the broadening of the waves, from a model consistent with experimental data
and analysed theoretically, in which cases such effects are dominant in practical situations.
Such analysis must be taken into account for implementing a possible correction of these
effects to be applicable in practice.
The reduction of amplitude of the signals leads to quantification errors in the reconstructed
distribution of the optical absorption coefficient and the broadening of the signals causes loss
of resolution of the images. It is important to take into account that both effects, quantification
errors and loss of resolution, can be due to other causes. Thus, the optical attenuation
also leads to quantification errors in the reconstructed images and its effect is usually more
important than the effect of acoustic attenuation. Only at high frequencies, the effect of
acoustic attenuation becomes dominant. On the other hand, the broadening of the signals
and the subsequent loss of resolution in the reconstructed images can be due to other causes
apart from acoustic attenuation. Specifically, any effect that limits the frequency content
of the signals causes broadening of the optoacoustic signals, for example the FR of the
transducer or the filters used to reduce the noise. Another effect that contributes to the loss
of resolution in tomographic setups is the space dependence of the speed of sound, which
6146 X L Deán-Ben et al

may be dominant in many cases. We have also shown theoretically that both the reduction of
amplitude and the broadening of the signals in water can be neglected with respect to tissue,
which is important to take into account as in most part of the setups the samples are immersed in
water. The two aforementioned effects of acoustic attenuation were also shown experimentally
for optoacoustic signals generated at a point source. The reduction of amplitude is clearly
perceived in the signals, and the frequency dependence of the attenuation coefficient can be
obtained by analysing the frequency components of the signals. The broadening of the signals
was also quantified. For low attenuation, the width of the signals is mainly due to the FR of
the transducer, whereas when the reduction of amplitude is large enough an increase in the
width of the signals is clearly perceived.
We presented a procedure to correct for the effects of acoustic attenuation in the
optoacoustic signals based on a model consistent with experimental measurements of the
acoustic attenuation. It is applicable in the time and in the frequency domains and, although
we only experimentally tested it for a uniform attenuating media, in principle it is applicable
for media with a space-dependent attenuation behaviour. It is important to take into account
that the validity of the attenuation model, in which we consider an attenuation coefficient
satisfying a power law dependence on frequency, must be verified in practice. Thus, the
frequency spectrum of the optoacoustic signals can contain very high frequencies, for which
this model has not been verified.
We also checked experimentally the performance of the suggested correction of the
acoustic attenuation effects. It is important to take into account that the quality of the
correction depends on the available frequencies in the spectrum of the attenuated signals.
Thus, for frequencies above a given value, the frequency spectrum is dominated by noise, so
that the attenuation correction leads to the unwanted amplification of the noise. Then, a low
pass filter must be applied to avoid this effect, which also limits the available frequency range
of the corrected signals. We have also shown that the results of the correction yielded by
assuming an attenuation coefficient satisfying a generic power law dependence on frequency
and by considering the phase of the signal collected with no attenuation (what corresponds to
a perfect correction in the phase) do not represent a significant improvement with respect to
the results obtained by assuming a linearly dependent attenuation coefficient and no correction
in the phase. This indicates the convenience of the latest simple approximation in practical
situations where an accurate measurement of the frequency dependence of the attenuation
coefficient and of the dispersion cannot be done.
The effects of acoustic attenuation in the signals relate to the quality of the reconstructed
images. Both quantification errors and loss of resolution were present in the tomographic
reconstruction of a phantom by placing an attenuating sample between the phantom and the
ultrasonic transducer. Both effects were partially reduced by applying the suggested correction
to the individual signals and then making the reconstruction, which indicates the feasibility of
correcting for attenuating effects in practical situations. It is important to take into account
that usually the attenuating material has a different speed of sound than water that must also be
considered. In the simple case that we tested experimentally, the correction of the difference
in the speed of sound was made by time-shifting the individual signals. The time-shifting is
constant in our case as the distance the ultrasonic waves propagate in the attenuating tissue
does not depend on the positions of the sources and the transducer. However, when the sample
itself consists of an attenuating material, the distance an ultrasonic wave propagates along the
attenuating medium is not constant, so that both the corrections for the acoustic attenuation
effects and the speed of sound variations become much more cumbersome in this case.
Finally, it is important to take into account that the material we used in the experiment
is much more attenuating than standard tissue. The attenuation coefficient is, however,
The effects of acoustic attenuation in optoacoustic signals 6147

approximately linear with frequency so the results obtained in the paper are equivalent to
the results obtained in standard tissue for much higher frequencies, provided that the linear
dependence of frequency of the attenuation coefficient is verified at such frequencies.
In conclusion, in this work we have shown both theoretical and experimentally the effects
of acoustic attenuation in optoacoustic signals. The theoretical analysis focused on the effects
of acoustic attenuation in practical situations by comparing them with effects caused by other
factors, so that we can identify in which cases the distortion of the experimental signals is
mainly due to acoustic attenuation. Thereby, we can assess in which cases it is convenient
to make a correction of its effects. The experimental results show the effects of acoustic
attenuation in actual cases, which fit the theoretical prediction. Also, the suggested correction
is applied both to individual signals and to a reconstructed image. The improvement of the
results obtained with the corrected signals indicates the feasibility of correcting for the effects
of acoustic attenuation in cases where they produce unwanted distortion in the images.

Acknowledgments

DR acknowledges support from the German Research Foundation (DFG) Research Grant (RA
1848/1) and the European Research Council (ERC) Starting Grant. VN acknowledges support
from the ERC Advanced Investigator Award and the German Ministry of Education (BMBF)
Innovation in Medicine Award.

References

Burgholzer P, Roitner H, Bauer-Marschallinger J and Paltauf G 2010 Image reconstruction in


photoacoustic tomography using integrating detectors accounting for frequency-dependent attenuation Proc.
SPIE 7564 75640O
Cox B T and Treeby B E 2010 Artifact trapping during time reversal photoacoustic imaging for acoustically
heterogeneous media IEEE Trans. Med. Imaging 29 387–96
Deán-Ben X L, Ma R, Razansky D and Ntziachristos V 2011a Statistical approach for optoacoustic image
reconstruction in the presence of strong acoustic heterogeneities IEEE Trans. Med. Imaging 30 401–8
Deán-Ben X L, Ntziachristos V and Razansky D 2011b Time-shifting correction in optoacoustic tomographic imaging
for media with a non-uniform speed of sound Proc. SPIE 8090 809013
Deán-Ben X L, Ntziachristos V and Razansky D 2011c Statistical optoacoustic image reconstruction using a priori
knowledge on the location of acoustic distortions Appl. Phys. Lett. 98 171110
Guo Z, S Hu and Wang L V 2010 Calibration-free absolute quantification of optical absorption coefficients using
acoustic spectra in 3D photoacoustic microscopy of biological tissue Opt. Lett. 35 2067–9
Gusev V E and Karabutov A A (ed) 1993 Laser Optoacoustics (New York: American Institute of Physics)
Hoelen C G A and de Mul F F M 2000 Image reconstruction for photoacoustic scanning of tissue structures Appl.
Opt. 39 5872–83
Jetzfellner T, Razansky D, Rosenthal A, Schulz R, Englmeier K H and Ntziachristos V 2009 Performance of iterative
optoacoustic tomography with experimental data Appl. Phys. Lett. 95 013703
Jiang H, Yuan Z and Gu X 2006 Spatially varying optical and acoustic property reconstruction using finite-element-
based photoacoustic tomography J. Opt. Soc. Am. A 23 878–88
Jose J, Willemink R G H, Resink S, Piras D, van Hespen J C G, Slump C H, Steenbergen W, van Leeuwen
T G and Manohar S 2011 Passive element enriched photoacoustic computed tomography (PER PACT) for
simultaneous imaging of acoustic propagation properties and light absorption Opt. Express 19 2093–104
Köstli K P, Frauchiger D, Niederhauser J J and Paltauf G 2001 Optoacoustic imaging using a three-dimensional
reconstruction algorithm IEEE J. Sel. Top. Quantum Electron. 7 918–23
La Riviere P J, Zhang J and Anastasio M A 2005 Image reconstruction in optoacoustic tomography accounting for
frequency-dependent attenuation IEEE Nucl. Sci. Symp. Conf. Rec. M03-337 1841–5
La Riviere P J, Zhang J and Anastasio M A 2006 Image reconstruction in optoacoustic tomography for dispersive
acoustic media Opt. Lett. 31 781–3
Ma R, Taruttis A, Ntziachristos V and Razansky D 2009 Multispectral optoacoustic tomography (MSOT) scanner for
whole-body small animal imaging Opt. Express 17 21414–26
6148 X L Deán-Ben et al

Ntziachristos V and Razansky D 2010 Molecular imaging by means of multispectral optoacoustic tomography
(MSOT) Chem. Rev. 110 2783–94
Nussenzveig H M (ed) 1972 Causality and Dispersion Relations (New York: Academic)
O’Donell M, Jaynes E T and Miller J G 1978 General relationships between ultrasonic attenuation and dispersion
J. Acoust. Soc. Am. 63 1935–7
Patch S K and Greenleaf A 2007 Ultrasound attenuation and thermo/photo/opto-acoustic tomography. Theoretical
foundation Proc. SPIE 6437 643726
Razansky D and Ntziachristos V 2007 Hybrid photoacoustic fluorescence molecular tomography using finite-element-
based inversion Med. Phys. 34 4293–301
Razansky D, Distel M, Vinegoni C, R Ma, Perrimon N, Köster R W and Ntziachristos V 2009 Multispectral opto-
acoustic tomography of deep-seated fluorescent proteins in vivo Nat. Photonics 3 412–7
Sushilov N V and Cobbold R S C 2004 Frequency-domain wave equation and its time-domain solutions in attenuating
media J. Acoust. Soc. Am. 115 1431–6
Szabo T L (ed) 2004 Diagnostic Ultrasound Imaging: Inside Out (San Diego, USA: Elsevier)
Treeby B E and Cox B T 2009 Fast, tissue-realistic models of photoacoustic wave propagation for homogeneous
attenuating media Proc. SPIE 7177 717716
Treeby B E and Cox B T 2010 Modeling power law absorption and dispersion for acoustic propagation using the
fractional Laplacian J. Acoust. Soc. Am. 127 2741–8
Treeby B E, Zhang E Z and Cox B T 2010 Photoacoustic tomography in absorbing acoustic media using time reversal
Inverse Problems 26 115003
Wang L V (ed) 2009 Photoacoustic Imaging and Spectroscopy (Boca Raton, USA: CRC Press)
Waters K R, Mobley J and Miller J G 2005 Causality-imposed (Kramers–Kronig) relationships between attenuation
and dispersion IEEE Trans. Ultrason. Ferroelectr. Freq. Control 52 822–33
Wells P N T 1999 Ultrasonic imaging of the human body Rep. Prog. Phys. 62 671–722
Xu Y and Wang L V 2003 Effects of acoustic heterogeneity in breast thermoacoustic tomography IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 50 1134–46
Xu M and Wang L V 2006 Photoacoustic imaging in biomedicine Rev. Sci. Instrum. 77 041101
Yuan Z, Wang Q and Jiang H 2007 Reconstruction of optical absorption coefficient maps of heterogeneous
media by photoacoustic tomography coupled with diffusion equation based regularized Newton method Opt.
Express 15 18076–81
Zhang C and Wang Y 2008 A reconstruction algorithm for thermoacoustic tomography with compensation for acoustic
speed heterogeneity Phys. Med. Biol. 53 4971–82

You might also like