Cone-beam artifact may be observed in the images reconstructed from circular trajectory data by use of the FDK algorithm or its variants for an imaged subject with longitudinally strong contrast variation in advanced diagnostic CT with a large number of detector rows. Existing algorithms have limited success in correcting for the effect of the cone-beam artifacts especially on the reconstruction of low-contrast soft-tissue. In the work, we investigate and develop optimization-based reconstruction algorithms to compensate for the cone-beam artifacts in the reconstruction of low-contrast anatomies. Specifically, we investigate the impact of optimization-based reconstruction design based upon different data-fidelity terms on the artifact correction by using the Chambolle- Pock (CP) algorithm tailored to each of the specific data-fidelity terms considered. We performed numerical studies with real data collected with the 320-slice Canon Medical System CT scanner, demonstrated the effectiveness of the optimization-based reconstruction design, and identified the optimization-based reconstruction that corrects most effectively for the cone-beam artifacts.
Markov random field (MRF) model has been widely used in Bayesian image reconstruction to reconstruct piecewise smooth images in the presence of noise, such as in low-dose X-ray computed tomography (LdCT). While it can preserve edge sharpness via edge-preserving potential function, its regional smoothing may sacrifice tissue image textures, which have been recognized as useful imaging biomarkers, and thus it compromises clinical tasks such as differentiating malignant vs. benign lesions, e.g., lung nodule or colon polyp. This study aims to shift the edge preserving regional noise smoothing paradigm to texture-preserving framework for LdCT image reconstruction while retaining the advantage of MRF’s neighborhood system on edge preservation. Specifically, we adapted the MRF model to incorporate the image textures of lung, bone, fat, muscle, etc. from previous full-dose CT scan as a priori knowledge for texture-preserving Bayesian reconstruction of current LdCT images. To show the feasibility of proposed reconstruction framework, experiments using clinical patient scans (with lung nodule or colon polyp) were conducted. The experimental outcomes showed noticeable gain by the a priori knowledge for LdCT image reconstruction with the well-known Haralick texture measures. Thus, it is conjectured that texture-preserving LdCT reconstruction has advantages over edge-preserving regional smoothing paradigm for texture-specific clinical applications.
Signal sparsity in computed tomography (CT) image reconstruction field is routinely interpreted as sparse angular sampling around the patient body whose image is to be reconstructed. For CT clinical applications, while the normal tissues may be known and treated as sparse signals but the abnormalities inside the body are usually unknown signals and may not be treated as sparse signals. Furthermore, the locations and structures of abnormalities are also usually unknown, and this uncertainty adds in more challenges in interpreting signal sparsity for clinical applications. In this exploratory experimental study, we assume that once the projection data around the continuous body are discretized regardless at what sampling rate, the image reconstruction of the continuous body from the discretized data becomes a signal sparse problem. We hypothesize that a dense prior model describing the continuous body is a desirable choice for achieving an optimal solution for a given clinical task. We tested this hypothesis by adapting total variation stroke (TVS) model to describe the continuous body signals and showing the gain over the classic filtered backprojection (FBP) at a wide range of angular sampling rate. For the given clinical task of detecting lung nodules of size 5mm and larger, a consistent improvement of TVS over FBP on nodule detection was observed by an experienced radiologists from low sample rate to high sampling rate. This experimental outcome concurs with the expectation of the TVS model. Further investigation for theoretical insights and task-dependent evaluations is needed.
To reduce radiation dose in X-ray computed tomography (CT) imaging, one of the common strategies is to lower the milliampere-second (mAs) setting during projection data acquisition. However, this strategy would inevitably increase the projection data noise, and the resulting image by the filtered back-projection (FBP) method may suffer from excessive noise and streak artifacts. The edge-preserving nonlocal means (NLM) filtering can help to reduce the noise-induced artifacts in the FBP reconstructed image, but it sometimes cannot completely eliminate them, especially under very low-dose circumstance when the image is severely degraded. To deal with this situation, we proposed a statistical image reconstruction scheme using a NLM-based regularization, which can suppress the noise and streak artifacts more effectively. However, we noticed that using uniform filtering parameter in the NLM-based regularization was rarely optimal for the entire image. Therefore, in this study, we further developed a novel approach for designing adaptive filtering parameters by considering local characteristics of the image, and the resulting regularization is referred to as adaptive NLM-based regularization. Experimental results with physical phantom and clinical patient data validated the superiority of using the proposed adaptive NLM-regularized statistical image reconstruction method for low-dose X-ray CT, in terms of noise/streak artifacts suppression and edge/detail/contrast/texture preservation.
Dual-energy computed tomography (DECT) is a recent advancement in CT technology, which can potentially reduce artifacts and provide accurate quantitative information for diagnosis. Recently, statistical iterative reconstruction (SIR) methods were introduced to DECT for radiation dose reduction. The statistical noise modeling of measurement data plays an important role in SIR and impacts on the image quality. Contrary to the conventional CT projection data, of which noise is independent from ray to ray, in spectral CT the basis material sinogram data has strong correlations. In order to analyze the image quality improvement by applying correlated noise model, we compare the effects of two different noise models (i.e., correlated noise model and independent model by ignoring correlations) by analyzing the bias and variance trade-off. The results indicate that in the same bias level, the correlated noise modeling results in up to 20.02% noise reduction compared to the independent noise model. In addition, their impacts to different numerical are also evaluated. The results show that using the non-diagonal covariance matrix in SIR is challenging, where some numerical algorithms such as a direct application of separable paraboloidal surrogates (SPS) cannot converge to the correct results.
One hundred “normal-dose” computed tomography (CT) studies of the chest (i.e., 1,160 projection views, 120kVp, 100mAs) data sets were acquired from the patients who were scheduled for lung biopsy at Stony Brook University Hospital under informed consent approved by our Institutional Review Board. To mimic low-dose CT imaging scenario (i.e., sparse-view scan), sparse projection views were evenly extracted from the total 1,160 projections of each patient and the total radiation dose was reduced according to how many sparse views were selected. A standard filtered backprojection (FBP) algorithm was applied to the 1160 projections to produce reference images for comparison purpose. In the low-dose scenario, both the FBP and total variation-stokes (TVS) algorithms were applied to reconstruct the corresponding low-dose images. The reconstructed images were evaluated by an experienced thoracic radiologist against the reference images. Both the low-dose reconstructions and the reference images were displayed on a 4- megapixel monitor in soft tissue and lung windows. The images were graded by a five-point scale from 0 to 4 (0, nondiagnostic; 1, severe artifact with low confidence; 2, moderate artifact or moderate diagnostic confidences; 3, mild artifact or high confidence; 4, well depicted without artifacts). Quantitative evaluation measurements such as standard deviations for different tissue types and universal quality index were also studied and reported for the results. The evaluation concluded that the TVS can reduce the view number from 1,160 to 580 with slightly lower scores as the reference, resulting in a dose reduction to close 50%.
Statistical iterative reconstruction (SIR) methods have shown remarkable gains over the conventional filtered
backprojection (FBP) method in improving image quality for low-dose computed tomography (CT). They reconstruct
the CT images by maximizing/minimizing a cost function in a statistical sense, where the cost function usually consists
of two terms: the data-fidelity term modeling the statistics of measured data, and the regularization term reflecting a
prior information. The regularization term in SIR plays a critical role for successful image reconstruction, and an
established family of regularizations is based on the Markov random field (MRF) model. Inspired by the success of nonlocal
means (NLM) algorithm in image processing applications, we proposed, in this work, a family of generic and edgepreserving
NLM-based regularizations for SIR. We evaluated one of them where the potential function takes the
quadratic-form. Experimental results with both digital and physical phantoms clearly demonstrated that SIR with the
proposed regularization can achieve more significant gains than SIR with the widely-used Gaussian MRF regularization
and the conventional FBP method, in terms of image noise reduction and resolution preservation.
Previous studies have reported that the volume-weighting technique has advantages over the linear interpolation
technique for cone-beam computed tomography (CBCT) image reconstruction. However, directly calculating the
intersecting volume between the pencil beam X-ray and the object is a challenge due to the computational complexity.
Inspired by previous works in area-simulating volume (ASV) technique for 3D positron emission tomography, we
proposed an improved ASV (IASV) technique, which can fast calculate the geometric probability of the intersection
between the pencil beam and the object. In order to show the improvements of using IASV technique in volumeweighting
based Feldkamp–Davis–Kress (VW-FDK) algorithm compared to the conventional linear interpolation
technique based FDK algorithm (LI-FDK), the variances images from both theoretical prediction and empirical
determination are described basing on the assumption of the uncorrelated and stationary noise for each detector bin. In
digital phantom study, both of the theoretically predicted variance images and the empirically determined variance
images concurred and demonstrated that the VW-FDK algorithm can result in uniformly distributed noise across the
FOV. In the physical phantom study, the performance enhancements by the VW-FDK algorithm were quantitatively
evaluated by the contrast-noise-ratio (CNR) merit. The CNR values from the VW-FDK result were about 40% higher
than the conventional LI-FDK result. Therefore it can be concluded that the VW-FDK algorithm can efficiently address
the non-uniformity noise and suppress noise level of the reconstructed images.
Cone-beam computed tomography (CBCT) has attracted growing interest of researchers in image reconstruction. The
mAs level of the X-ray tube current, in practical application of CBCT, is mitigated in order to reduce the CBCT dose.
The lowering of the X-ray tube current, however, results in the degradation of image quality. Thus, low-dose CBCT
image reconstruction is in effect a noise problem. To acquire clinically acceptable quality of image, and keep the X-ray
tube current as low as achievable in the meanwhile, some penalized weighted least-squares (PWLS)-based image
reconstruction algorithms have been developed. One representative strategy in previous work is to model the prior
information for solution regularization using an anisotropic penalty term. To enhance the edge preserving and noise
suppressing in a finer scale, a novel algorithm combining the local binary pattern (LBP) with penalized weighted leastsquares
(PWLS), called LBP-PWLS-based image reconstruction algorithm, is proposed in this work. The proposed
LBP-PWLS-based algorithm adaptively encourages strong diffusion on the local spot/flat region around a voxel and less
diffusion on edge/corner ones by adjusting the penalty for cost function, after the LBP is utilized to detect the region
around the voxel as spot, flat and edge ones. The LBP-PWLS-based reconstruction algorithm was evaluated using the
sinogram data acquired by a clinical CT scanner from the CatPhan® 600 phantom. Experimental results on the noiseresolution
tradeoff measurement and other quantitative measurements demonstrated its feasibility and effectiveness in
edge preserving and noise suppressing in comparison with a previous PWLS reconstruction algorithm.
Reducing X-ray exposure to the patients is one of the major research efforts in the computed tomography (CT) field, and
one of the common strategies to achieve it is to lower the mAs setting (by lowering the X-ray tube current and/or
shortening the exposure time) in currently available CT scanners. However, the image quality from low mAs acquisition
is severely degraded due to excessive quantum noise, if no adequate noise control is applied during image
reconstruction. Different from filter-based algorithms, statistical reconstruction algorithms model the statistical property
of the noise using a cost function and minimize the cost function for an optimal solution in statistical sense. The
algorithms have shown to be feasible and effective in both sinogram and image domain. In our previous researches, we
proposed penalized reweighted least-squares (PRWLS) approaches to sinogram noise reduction and image
reconstruction for low-dose CT imaging, which are in this statistical category. This work is a continuation of the
research along this direction and aims to compare the reconstruction quality of two different PRWLS implementations
for low-dose cone-beam CT reconstruction: (1) PRWLS sinogram restoration followed by analytical Feldkamp-Davis-
Kress reconstruction, (2) fully iterative PRWLS image reconstruction. Inspired by our recent study on the variance of
low-mAs projection data in presence of electric noise background, a more accurate weight was adopted in the weighted
least-squares term. An anisotropic quadratic form penalty was utilized in both PRWLS implementations to preserve
edges during noise reduction. Experiments using the CatPhan® 600 phantom and anthropomorphic head phantom were
carried to study the relevant performance of these two implementations on image reconstruction. The results revealed
that the implementation (2) can outperform implementation (1) in terms of noise-resolution tradeoff measurement and
analysis of the reconstructed small objects due to its matched image edge-preserved penalty in the image domain.
However, those gains are offset by the cost of increased computational time. Thus, further examination of real patient
data is necessary to show the clinical significance of the iterative PRWLS image reconstruction over the PRWLS
sinogram restoration.
This paper introduces a new strategy to reconstruct computed tomography (CT) images from sparse-view projection data
based on total variation stokes (TVS) strategy. Previous works have shown that CT images can be reconstructed from
sparse-view data by solving a constrained TV problem. Considering the incompressible property of the voxels along the
tangent direction of isophote lines, a tangent vector is consolidated in this newly-proposed algorithm for normal vector
estimation. Then, a minimization problem based on this estimated normal vector is addressed and resolved in
computation. The to-be-estimated image is obtained by executing this two-step framework iteratively with projection
data fidelity constraints. By introducing this normal vector estimation, the edge information of the image is well
preserved and the artifacts are efficiently inhibited. In addition, the new proposed algorithm can mitigate the staircase
effects which are usually observed from the results of the conventional constrained TV method. In this study, the TVS
method was evaluated by patients’ brain raw data which was acquired from Siemens SOMATOM Sensation 16-slice CT
scanner. The results suggest that the proposed TVS strategy can accurately reconstruct the brain images and produce
comparable results relative to the TV-projection onto convex sets (TV-POCS) method and its general case: adaptiveweighted
TV-POCS (AwTV-POCS) method from 232,116 projection views. In addition, an improvement was observed
when using only 77 views for TVS method compared to the AwTV/TV-POCS methods. In the quantitative evaluation,
the TVS method showed adequate noise-resolution property and highest universal quality index value.
Helical computed tomography (HCT) has demonstrated the effectiveness in virtual colonoscopy (VC) or CTcolonography
(CTC). One major concern with this clinical application is associated with the risk of high radiation
exposure, especially for its use for screening purpose at a large population. In this work, we presented an improved
Karhunen-Loeve (KL) domain penalized weighted least-squares (PWLS) strategy which considers the data correlations
among the projection rays mainly due to partially overlap while system rotates. Two 1-dimensional (1D) projections,
which called coupled projections (CPs), are composed according to the geometry. Each element of the 1D projection is
carefully selected for a specific point within 2π angle along the system rotates and thus a highly correlation can be
observed between any specific projection and the CPs. These highly correlated projections can be treated by an adaptive
KL-PWLS strategy for accurate noise reduction. This method has been implemented and tested on computer simulated
sinograms which mimic low-dose CT scans. The reconstructed images by the presented strategy demonstrated the
potential of ultra low-dose CT application.
This paper introduced a novel method to reconstruct low-dose computed tomography (LDCT) images based on weighted
total variation minimization. Previous work has shown that a CT image can be reconstructed from under-sampling data
by minimizing the total variation (TV) of the image with some constraints. However, due to the piecewise constant
assumption, the conventional TV minimization algorithm often suffers over-smooth on the edges of the resulted image.
Considering the anisotropic property of the edge voxels in an image, we proposed a weighted total variation (WTV)
minimization algorithm to achieve high-quality LDCT image reconstruction from under-sampling sinogram data. The
proposed algorithm was evaluated by computer simulation, where the Shepp-Logan phantom was used. In noise free
cases, 20 projection views were enough to accurately reconstruct the phantom image by the WTV approach while the
filtered backprojection (FBP) reconstruction failed. In noise cases, different levels of Poisson noise were added to the
noise-free sinogram. The local signal to noise ratio (SNR) and global SNR were computed to evaluate the robustness of
the WTV algorithm to the noise. The simulation results showed that the WTV algorithm can have much higher global
SNR compared to the conventional TV. The relationship between the number of views of WTV algorithm and the mean
square error (MSE) of the reconstructed images was also discussed in this paper in order to find the minimum projection
numbers required to obtain an adequate reconstruction image for
low-dose CT applications.
Low-dose x-ray computed tomography (CT) is clinically desired. However, the quality of low-dose CT image is
severely degraded due to excessive photon quantum noise and electronic noise. It is known that accurate noise modeling
is a fundamental issue for low-dose CT imaging, such as statistical iterative image reconstruction and statistics-based
singoram restoration. In this paper, we first studied the statistical moment properties of the noise model in CT projection
domain wherein the noise of detected signal is considered as quantum photon noise plus background electronic noise.
More importantly, we derived a new formula to estimate the mean-variance relationship in Radon domain by using
Taylor explanation. To test the presented variance estimation formula, an anthropomorphic torso phantom was scanned
repeated by a commercial scanner at five different mAs levels from 100 down to 17. The experimental results
demonstrate that the electronic noise is significant when low-dose scan is performed or the number of detected photons
is limited. As an important conclusion of the presented study, electronic noise effect should be considered in low-dose
CT image reconstruction.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.