View PDF

Many-view under-sampling (MVUS)
technique for low-dose CT
Taewon Lee1, Sajid Abbas1, Byungchul Cho2, Insoo Kim3, Bumsoo Han3, and Seungryong Cho1*
Abstract–In computed tomography (CT) imaging, radiation
dose delivered to the patient is one of the major concerns. Among
many technical solutions to lowering radiation dose while
preserving clinical utilities of the images, sparse-view CT is
promising technique. However, a fast power switching of an Xray tube, which is needed for the sparse-view sampling, can be
challenging in many CT systems. We have recently proposed a
novel alternative approach to sparse-view circular CT that can
be readily incorporated on the existing CT systems, and have
successfully shown its feasibility. Instead of switching the X-ray
tube power, one can place an oscillating multi-slit collimator
between the X-ray tube and the patient to partially block the Xray beam thereby reducing the radiation. In this study, an
experimental study was performed to evaluate the performance
of the proposed XT scan scheme. Industrial CT projection data
of a CatPhan® 600 phantom was acquired by use of the
oscillating multi-slit collimator. We used a sinusoidal motion of
the collimator to the perpendicular direction of the rotation axis
for the purpose of obtaining more uniform spatial sampling of
the image. For image reconstruction, we used a total-variation
minimization (TV) algorithm which has shown its outperformance in many sparse-view CT applications.
a kind of sparse-view CT is implemented as shown in Fig. 1.
This technique has pros and cons. The X-ray scatter by the
patient can be reduced, and even more corrected if needed by
use of the shadow of the beam block strips in the collimator.
Reduction of scatter is thought to contribute to enhancing the
consistency of the data to the imaging model, thereby
potentially useful for improving image quality with possibly a
better convergence in the iterative image reconstruction.
Additionally, if the beam blockers are partly transparent to the
X-ray, then a dual-energy imaging would be relatively easily
achieved by use of the different energy spectra of X-ray in a
single scan. The cons include the penumbra effect and the
motion artifact due to the collimator oscillation in the
projection data, which however may be corrected for.
I. INTRODUCTION
been widely used for many clinical applications.
CTDosehasreduction
seems to be the most important issue to the
CT developers and researchers. Among many technical
solutions to lower radiation dose while preserving clinical
utilities of the images, sparse-view CT is a promising
technique. Sparse-view CT, which takes fewer projections,
provides a viable option to reducing radiation dose. However
it is technologically hard to implement in the current CT
systems due to difficulties in fast tube power switching. We
have recently proposed a many-view under sampling (MVUS)
techniques as an alternative to sparse-view CT [1]. Instead of
switching the X-ray tube power, a multi-slit collimator is
placed between the X-ray tube and patient to partially block
the X-ray beam thereby reducing the radiation. A sparsesampling CT is realized by placing the oscillation multi-slit
collimator between the X-ray tube and the patient. As a result,
The work was supported in part by the NRF grant NRF2013M2A2A9043476, and by the MEST grant R0001270 and R0001376 in
Korea.
T. Lee, S. Abbas, S. Cho are with the Korea Advanced Institute of Science
and Technology, Daejeon, South Korea (telephone: +82 42-350-3828, email:[email protected]).
B. Cho is with Asan Medical Center, University of Ulsan College of
Medicine, Seoul, South Korea.
I. Kim, B. Han are with EB Tech Co., Ltd. Yongsan-dong 550 Yuseonggu, Daejeon, 305-500, South Korea.
978-1-4799-0534-8/13/$31.00 ©2013 IEEE
Fig. 1. Schematic of the proposed scanning configuration is
illustrated. The arrow indicates a reciprocating motion of the
collimator.
II. MATERIAL AND METHOD
A. Industrial CT system
We used an Industrial cone-beam CT projection data set of
a CatPhan® 600, which includes 720 projections per rotation
and realized MVUS scanning by using an oscillating multi-slit
collimator. A detailed description of the experiment is given
below with the image reconstruction. Scanning parameters
used in the data acquisition are summarized in table 1. We
also fabricated a multi-slit collimator as shown in Fig. 2. A
multi-slit collimator was made of a tungsten thin plate. The
collimator is composed periodic arrangement of slit-opening
and radio-opaque rectangular area, and the length dimension
of the slits is parallel to the rotation axis.
TABLE I. SCANNING CONDITIONS
Parameter
Tube voltage
Tube current
Value
120kVp
1.2mA/ 4.8mA
Detector size
1024 x 1024, 400 ㎛
Scan range
Number of projections
Source to object distance
Source to detector distance
360°
720
800mm
1500mm
Fig. 3. The sonograms are shown corresponding to (a) a full
data sampling case, (b) one-fourth of the data used for image
reconstruction according to the MVUS technique.
Fig. 2. The CT system with the collimator mounted is
shown in (a).
B. Data sampling scheme
In our earlier work, various sparse samplinng schemes were
investigated numerically [4]. We recruited two measures in the
context of compress sensing theory: sampling density (SD),
data incoherence (DIC). Effects of SD, DIC on image quality
reconstructed by total-variation minimization algorithm were
studied, and several schemes were shown to provide good
quality of images compared to others.
Out of such candidate schemes for sparse sampling, we
particularly used a collimator-based sampling scheme and
implemented an oscillating motion of the collimator to the
perpendicular direction of the rotation axis for acquiring more
uniform spatial sampling and less correlated data acquisition.
For dose reduction, the size of slit-opening is one quarter to
the repetition pattern in the anticipation of dose reduction by a
factor of 3/4. The frequency of the motion was 30 trips per
scanner rotation. Figure 3 shows the sinograms comparatively
between a conventional scan and the proposed scan.
C. Total-Variation Minimization Algorithm
For image reconstruction, we used a total-variation
minimization (TV) algorithm. The TV algorithm is based on
the compressive sensing theory and its excellent performance
in sparse-view CT applications has been reported [2]. We
adopted the adaptive-steepest-decent projection-onto-convexsets (ASD-POCS) approach [3] and modified the POCS step
so that only the measured data through the collimator slits are
to be used in the computation.
The TV algorithm searches for a solution that minimizes the
image total-variation.
which satisfies two constrains,
and
0
can be selected for controlling the impact level of data
inconsistency on the image reconstruction.
the
,where f represents an image under iteration,
minimum image total-variation solution, M the system matrix,
and g the measured data.
represents the total-variation
of an image function. The system matrix was based on a raydriven model.
III. RESULTS
Fig. 4. shows the reconstruction images of a transverse slice
of CatPhan® 600. Image reconstructed by the FBP algorithm
from the low mAs uncollimated 720 projections is shown in
Fig. 4 (a) as a reference image. We used about 1/4 of the
standard tube current used in the system for a fair comparison.
Image reconstructed by use of the total-variation minimization
algorithm from the collimated 720 projections with a standard
mAs condition according to 1/4 collimation ratio is shown in
Fig. 4 (b).
REFERENCES
[1]
[2]
[3]
[4]
Fig.4. The reconstruction images of a CatPhan® 600 is
shown. The image was reconstructed (a) by the FBP algorithm
from low-dose 720 projections. The image was reconstructed
(b) one-fourth of the total area collimated 720 projections.
Additionally, we calculated an image contrast index, CNR
to quantitatively assess the image quality. The ROI was
selected as shown in Fig. 5. The corresponding CNR values
are summarized in Table 2.
TABLE Ⅱ. CNR VALUE OF EACH EXPERIMENT CONDITION
CNR
FBP
2.5
MVUS
11.1
Fig. 5. Selected ROI in result images.
IV. CONCLUSION
In our study, we have experimentally demonstrated the
feasibility of a novel sampling scheme for low-dose CT,
which is named MVUS. Real experiment with a moving
multi-slit collimator was conducted, and a preliminary image
reconstruction was successfully shown. More investigation in
under work for optimizing the MVUS moving scheme,
utilizing the SD & DIC information for improving
performance, investigating the effect of scattter reduction and
correction, task-dependent image quality asssessment, testing
dual-energy imaging feasibility, and analyzing and correcting
the penumbra.
S. Cho, T. Lee, J. Min, H. Chung, "Feasibility study on many-view
under sampling (MVUS) technique for low-dose computed tomography",
Opt. Eng., 51, 120837L, 2012.
E. Y. Sidky, C. M. Kao, and X. Pan, "Accurate image reconstruction
from few-views and limited-angle data in divergent-beam CT", J. X-ray
Sci. Tech., vol. 14, pp. 11_139, 2006.
E. Y. Sidky and X. Pan, "Image reconstruction in circular cone-beam
computed tomography by con-strained, total-variation minimization",
Phys. Med. Biol., vol. 53, pp. 4777_4807, 2008.
S. Abbas, T. Lee, S. Shin, R. Lee, and S. Cho, "Effects of sparse
sampling schemes on image quality in low-dose CT", Med. Phys. vol. 40,
PP. 111915, 2013.

Download Report