COL 12(11), 111703(2014) CHINESE OPTICS LETTERS November 10, 2014
1671-7694/2014/111703(5) 111703-1 © 2014 Chinese Optics Letters
Photoacoustic imaging (PAI) is a noninvasive medical
imaging technique that has a great potential for clinic
applications such as early tumor detection
[1,2]
, vessel
imaging
[3,4]
, and brain imaging
[5]
. PAI technique can pro-
vide higher contrast
[6]
and resolution than ultrasound
imaging and is more eective for imaging deeper struc-
ture compared with pure optical imaging. It combines
the strengths of optical and ultrasound imaging
[7]
.
In the practical use of PAI, tissues are illuminated
with short laser pulses, which result in the generation
of acoustic waves because of the photoacoustic eect.
In this letter, we are concerned about the computed
tomographic PAI in the imaging mode. The propa-
gated photoacoustic signals are detected by a scanning
ultrasound transducer or a transducer array. With the
knowledge of these sampling data, the optical absorp-
tion deposition within the tissue can be estimated by
employing an image reconstruction algorithm.
The key point of imaging quality in PAI is the
reconstruction algorithms. Xu et al. proposed the l-
tered back-projection algorithm
[8]
for PAI, which has
been widely used for its convenience. The deconvolu-
tion reconstruction algorithm proposed by Zhang et
al. has specic advantages under the circumstance of
limited-angle sampling and heterogeneous acoustic
medium
[9,10]
. The above-mentioned algorithms are the
analytical reconstruction methods, which have advan-
tages in the computational cost and implementation
convenience. However, the analytical algorithms fail
to be eective when the sampling points are sparse
and the sparse-view imaging system is very important
to reduce data acquaintance time. This drawback lim-
its the applications of the analytical algorithms and
impairs their performance.
There also exists PAI system which can image the
whole area with one laser exposure. These systems
High total variation-based method for sparse-view
photoacoustic reconstruction
Chen Zhang (张 晨)
1
and Yuanyuan Wang (汪源源)
1,2
1
Department of Electronic Engineering, Fudan University, Shanghai 200433, China
2
Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention,
Shanghai 200433, China
Corresponding author: yywang@fudan.edu.cn
Received June 25, 2014; accepted August 8, 2014; posted online October 28, 2014
We propose a novel method by combining the total variation (TV) with the high-degree TV (HDTV) to
improve the reconstruction quality of sparse-view sampling photoacoustic imaging (PAI). A weighing function
is adaptively updated in an iterative way to combine the solutions of the TV and HDTV minimizations.
The fast iterative shrinkage/thresholding algorithm is implemented to solve both the TV and the HDTV
minimizations with better convergence rate. Numerical results demonstrate the superiority and eiciency of
the proposed method on sparse-view PAI. In vitro experiments also illustrate that the method can be used
in practical sparse-view PAI.
OCIS codes: 110.5120, 100.3010, 170.3880, 170.5120.
doi: 10.3788/COL201412.111703.
usually have large amount of transducers around
the imaging area. With the help of sparse-view PAI
reconstruction method, the transducer amount can be
reduced. This reduction benets the system from two
main aspects. Firstly, the system is easier to maintain
in a lower level of system complexity. Secondly, this re-
duction can make the data scale much smaller. Besides
these two aspects, it is also worth mentioning that it
reduces the cost of the system. These aspects are very
important for further clinical applications. So it is very
important to develop a sparse-view imaging system.
In order to avoid these shortcomings, the model-based
iterative algorithms are developed faster in recent years.
The iterative algorithms can provide improvement in
image quality and noise robustness
[11]
. Among them,
algorithms that adopted the compressed sensing (CS)
theory perform best in sparse-view reconstruction
[12]
.
The total variation (TV) method is involved in the CS
theory. The TV-based iterative algorithms can recover
the images accurately from the sparse sampling data
in PAI
[13,14]
. But it has been shown that the TV-based
algorithm sometimes transforms the smooth area into
piecewise constants and fails to show some detailed
information.
In this letter, we propose a novel algorithm for sparse-
view PAI image reconstruction. The algorithm combines
the TV minimization with the high-degree TV (HDTV)
minimization. Our contributions are threefold. Firstly,
we include the HDTV minimization into the PAI recon-
struction. This combined method is able to avoid the
painting such as artifacts in smooth regions and inherits
edge preservation advantage of the standard TV. Sec-
ondly, we implement a weighting function to combine
the solutions of the TV and the HDTV minimizations.
This weighting function is adaptively updated. Thirdly,
we extend the fast iterative shrinkage/thresholding