A multi-purpose readout electronics based on the DPLMS digital filter has been developed for CdTe and CZT detectors for X-ray imaging applications. Different filter coefficients can be synthesized optimized either for high energy resolution at relatively low counting rate or for high rate photon-counting with reduced energy resolution. The effects of signal width constraints, sampling rate and length were numerical studied by Mento Carlo simulation with simple CRRC shaper input signals. The signal width constraint had minor effect and the ENC was only increased by 6.5% when the signal width was shortened down to 2 τc. The sampling rate and length depended on the characteristic time constants of both input and output signals. For simple CR-RC input signals, the minimum number of the filter coefficients was 12 with 10% increase in ENC when the output time constant was close to the input shaping time. A prototype readout electronics was developed for demonstration, using a previously designed analog front ASIC and a commercial ADC card. Two different DPLMS filters were successfully synthesized and applied for high resolution and high counting rate applications respectively. The readout electronics was also tested with a linear array CdTe detector. The energy resolutions of Am-241 59.5 keV peak were measured to be 6.41% in FWHM for the high resolution filter and to be 13.58% in FWHM for the high counting rate filter with 160 ns signal width constraint.
Spectral CT is attracting more and more attention in medicine, industrial nondestructive testing and security inspection field. Material decomposition is an important issue to a spectral CT to discriminate materials. Because of the spectrum overlap of energy channels, as well as the correlation of basis functions, it is well acknowledged that decomposition step in spectral CT imaging causes noise amplification and artifacts in component coefficient images. In this work, we propose materials decomposition via an optimization method to improve the quality of decomposed coefficient images. On the basis of general optimization problem, total variance minimization is constrained on coefficient images in our overall objective function with adjustable weights. We solve this constrained optimization problem under the framework of ADMM. Validation on both a numerical dental phantom in simulation and a real phantom of pig leg on a practical CT system using dual-energy imaging is executed. Both numerical and physical experiments give visually obvious better reconstructions than a general direct inverse method. SNR and SSIM are adopted to quantitatively evaluate the image quality of decomposed component coefficients. All results demonstrate that the TV-constrained decomposition method performs well in reducing noise without losing spatial resolution so that improving the image quality. The method can be easily incorporated into different types of spectral imaging modalities, as well as for cases with energy channels more than two.
Dual-energy CT (DECT) imaging has gained a lot of attenuation because of its capability to discriminate materials. We proposes a flexible DECT scan strategy which can be realized on a system with general X-ray sources and detectors. In order to lower dose and scanning time, our DECT acquires two projections data sets on two arcs of limited-angular coverage (one for each energy) respectively. Meanwhile, a certain number of rays from two data sets form conjugate sampling pairs. Our reconstruction method for such a DECT scan mainly tackles the consequent limited-angle problem. Using the idea of artificial neural network, we excavate the connection between projections at two different energies by constructing a relationship between the linear attenuation coefficient of the high energy and that of the low one. We use this relationship to cross-estimate missing projections and reconstruct attenuation images from an augmented data set including projections at views covered by itself (projections collected in scanning) and by the other energy (projections estimated) for each energy respectively. Validated by our numerical experiment on a dental phantom with rather complex structures, our DECT is effective in recovering small structures in severe limited-angle situations. This DECT scanning strategy can much broaden DECT design in reality.
In this article, we present an easy-to-implement Multi-energy CT scanning strategy and a corresponding reconstruction method, which facilitate spectral CT imaging by improving the data efficiency the number-of-energy- channel fold without introducing visible limited-angle artifacts caused by reducing projection views. Leveraging the structure coherence at different energies, we first pre-reconstruct a prior structure information image using projection data from all energy channels. Then, we perform a k-means clustering on the prior image to generate a sparse dictionary representation for the image, which severs as a structure information constraint. We com- bine this constraint with conventional compressed sensing method and proposed a new model which we referred as Joint Clustering Prior and Sparsity Regularization (CPSR). CPSR is a convex problem and we solve it by Alternating Direction Method of Multipliers (ADMM).
We verify our CPSR reconstruction method with a numerical simulation experiment. A dental phantom with complicate structures of teeth and soft tissues is used. X-ray beams from three spectra of different peak energies (120kVp, 90kVp, 60kVp) irradiate the phantom to form tri-energy projections. Projection data covering only 75◦ from each energy spectrum are collected for reconstruction. Independent reconstruction for each energy will cause severe limited-angle artifacts even with the help of compressed sensing approaches. Our CPSR provides us with images free of the limited-angle artifact. All edge details are well preserved in our experimental study.
Multi-segment straight-line trajectory computed tomography (CT) requires accurate movement between each segment
trajectories. In industrial applications for large objects, it is difficult and very cost-expensive for the mechanical and
control system to ensure the accurate movement during the rotation process between two segment trajectories. Hence, precisely measuring the movement would be an alternative choice to acquire an exact reconstruction. In this work, we proposes a new method to measure the movement by using invariant moment of attenuation distribution images. Its accuracy improves as the number of projection data increases. The method is validated by both numerical and practical experiments. Reasonable CT reconstructions are obtained for a two-segment linear trajectory CT with coarse mechanical movement control modeled in our experiments.
This work gives a new Compressed Sensing (CS) based Computed Tomography (CT) reconstruction method for limited angle problem. Currently CS based reconstruction methods are achieved by a minimizing process on the total variation (TV) of CT image under data consistency constraint. For limited-angle problem due to the missing range of projection views the strength of data consistency constraint becomes direction relevant. In our work a new anisotropic total variation (ATV) minimization method is proposed. Instead of using image TV as the minimization objective, an ATV objective is designed which is combined of multiple 1D directional TV with different weights according to the actual scanned angular range. Experiments with simulated data demonstrate the advantages of our approach relative to the standard CS based reconstruction methods.
The Poisson-like noise model has been widely used for noise suppression and image reconstruction in low dose computed tomography. Various noise estimation and suppression approaches have been developed and studied to enhance the image quality. Among them, the recently proposed generalized Anscombe transform (GAT) has been utilized to stabilize the variance of Poisson-Gaussian noise. In this paper, we present a variance estimation approach using GAT. After the transform, the projection data is denoised conventionally with an assumption that the noise variance is uniformly equals to 1. The difference of the original and the denoised projection is treated as pure noise and the global variance σ2 can be estimated from the residual difference. Thus, the final denoising step with the estimated σ2 is performed. The proposed approach is verified on a cone-beam CT system and demonstrated to obtain a more accurate estimation of the actual parameter. We also examine FBP algorithm with the two-step noise suppression in the projection domain using the estimated noise variance. Reconstruction results with simulated and practical projection data suggest that the presented approach could be effective in practical imaging applications.
Helical CT scan has been acknowledged to be a very useful scanning mode. Normally, the speed of bed movement per
rotation (pitch) for a helical CT is fixed to meet the requirement of CT scanning speed. To reduce system cost, single
slice helical CT (SSHCT) is often chosen in many applications. It is interesting and useful in real life to answer the
question that how to design the detector to obtain optimal performance of a SSHCT in a detection task. In this work, we
applied ROC study for the optimization of detector thickness along the direction of rotation axis for our SSHCT.
Numerical simulations followed by human observer studies are done in this investigation. Compound Gaussian noises
are modeled in our numerical simulations for objects both with and without lesions. An analytical FBP reconstruction
method with rebinning is used for noisy data reconstruction. It can be seen in the reconstructions that thin detectors lead
to artifacts, and that thick detectors lead to lesion blurring and lower contrast. All these impact on lesion detection in the
practical imaging applications. According to our ROC tests done on images from five choices of detector thickness,
optimal performance is obtained when choosing detector thickness being around 1~1.25 times of the helical pitch.
Moreover, we find that, under different noise level, the optimal point is about the same.
Some other figures of merit including SNR and HTC are also calculated and examined in this work. The results relates
well with the results of AUCs. It shows that, they could serve very well as the indicators for system optimization when
few non-linear physical effect and reconstruction processing are involved.
The aim of the present study is to investigate a type of Bayesian reconstruction which utilizes partial differential
equations (PDE) image models as regularization. PDE image models are widely used in image restoration and
segmentation. In a PDE model, the image can be viewed as the solution of an evolutionary differential equation. The
variation of the image can be regard as a descent of an energy function, which entitles us to use PDE models in Bayesian
reconstruction. In this paper, two PDE models called anisotropic diffusion are studied. Both of them have the
characteristics of edge-preserving and denoising like the popular median root prior (MRP). We use PDE regularization
with an Ordered Subsets accelerated Bayesian one step late (OSL) reconstruction algorithm for emission tomography.
The OS accelerated OSL algorithm is more practical than a non-accelerated one. The proposed algorithm is called
OSEM-PDE. We validated the OSEM-PDE using a Zubal phantom in numerical experiments with attenuation correction
and quantum noise considered, and the results are compared with OSEM and an OS version of MRP (OSEM-MRP)
reconstruction. OSEM-PDE shows better results both in bias and variance. The reconstruction images are smoother and
have sharper edges, thus are more applicable for post processing such as segmentation. We validate this using a k-means
segmentation algorithm. The classic OSEM is not convergent especially in noisy condition. However, in our experiment,
OSEM-PDE can benefit from OS acceleration and keep stable and convergent while OSEM-MRP failed to converge.
Helical cone-beam CT is widely used nowadays because of its rapid scan speed and efficient utilization of x-ray dose.
HCT-FDK is an effective reconstruction algorithm on Helical CT. However, like other 3D reconstruction algorithms,
HCT-FDK is time consuming because of its large amount of data processing including the convolution and 3D-3D back
projection. Recently, GPU is widely used to parallel many reconstruction algorithms. The latest GPU has some nice
features, such as large memory, lots of processors, fast 3D texture mapping, and flexible frame buffer object. All these
features help reconstruction a lot. In this paper, we present a solution to this problem with GPU. First, we bring a lookup
table into HCT-FDK. Then, both convolution and back projection are implemented on GPU. At last, the reconstruction
result is directly smoothed and visualized by GPU. Experimental results are given to compare among CPU and two
generations of GPU: Geforce 6800GT and Geforce 8800GTX. The comparison was applied both on simulation data and
real data. We show that, GPU-accelerated HCT-FDK gets result with similar levels of noise and clarity but gains a speed
increase of about 10-100 times faster than using CPU only. With its newer feature, Geforce 8800GTX can get a similar
quality like Geforce 6800GT and about 20 times faster.
Cosmic ray muon radiography which has a good penetrability and sensitivity to high-Z materials is an effective way for
detecting shielded nuclear materials. Reconstruction algorithm is the key point of this technique. Currently, there are two
main algorithms about this technique. One is the Point of Closest Approach (POCA) reconstruction algorithm which
uses the track information to reconstruct; the other is the Maximum Likelihood estimation, such as the Maximum
Likelihood Scattering (MLS) and the Maximum Likelihood Scattering and Displacement (MLSD) reconstruction
algorithms which are proposed by the Los Alamos National Laboratory (LANL). The performance of MLSD is better
than MLS. Since MLSD reconstruction algorithm includes scattering and displacement information while MLS reconstruction algorithm only includes scattering information. In order to get this Maximum Likelihood estimation, in this paper, we propose to use EM method to get the estimation (MLS-EM and MLSD-EM). Then, in order to saving reconstruction time we use the OS technique to accelerate MLS and MLSD reconstruction algorithm with the initial value set to be the result of the POCA reconstruction algorithm. That is, the Maximum Likelihood Scattering-OSEM (MLS-OSEM) and the Maximum Likelihood Scattering and Displacement-OSEM (MLSD-OSEM). Numerical simulations show that the MLSD-OSEM is an effective algorithm and the performance of MLSD-OSEM is better than MLS-OSEM.
In today's tomographic imaging, there are more incomplete data systems, such as few-view system. The advantage
of few-view tomography is less x-ray dose and reduced scanning time. In this work, we study the projection
distribution in few-view fan-beam imaging. It is one of the fundamental problems in few-view imaging because
of its severe lack of projection data. The aim is to reduce data redundancy and to improve the quality of reconstructed
images by research on projection distribution schemes. The reconstruction algorithm for few-view
imaging is based on algebraic reconstruction techniques (ART) and total variation (TV) constraint approached
by E. Sidky et al in 2006. Study of few-view fan-beam projection distribution is performed mainly through
comparison of several distribution types in projection space and reconstructed images. Results show that the
distribution called short-scan type obtains the best image in five typical distributions.
A new imaging configuration whose trajectory is a multisegment straight line is investigated, and a practical reconstruction algorithm is proposed. It is a natural extension of an imaging configuration with a straight-line trajectory. These kinds of scanning systems may be useful in industry and security inspections. As is known, projection data from a single straight-line trajectory are incomplete and their reconstruction suffers from a limited-angle problem. A multisegment straight-line trajectory can be used to compensate for this deficiency. To reconstruct images, a practical reconstruction algorithm is derived. It is of the Feldkamp-Davis-Kress (FDK) type, and is efficient and straightforward. Like the FDK algorithm, our reconstruction is exact in the midplane and can be exact everywhere if the density of the scanned object is independent of the direction z, though the integral of the reconstructed image along z is no longer preserved. Numerical simulations validate our method.
A computed tomography (CT) imaging configuration with a straight-line trajectory is investigated, and a direct filtered backprojection (FBP) algorithm is presented. This kind of system may be useful for industrial applications and security inspections. Projections from a straight-line trajectory have a special property where data from each detector element correspond to a parallel-beam projection of a certain view angle. However, the sampling steps of parallel beams differ from view to view. Rebinning raw projections into uniformly sampled parallel-beam projections is a common choice for this type of reconstruction problem. However, the rebinning procedure suffers a loss of spatial resolution because of interpolations. Our reconstruction method is first derived from the Fourier slice theorem, where a coordinate transform and geometrical relations in projection and backprojection are used. It is then extended to 3-D scanning geometry. Finally, data-shift preprocessing is proposed to reduce computation and memory requirements by removing useless projections in raw data. In this method, the spatial resolution is better preserved and the reconstruction is less sensitive to data truncation than in the rebinning-to-parallel-beam method. To deal with limited angle problem, an iterative reconstruction reprojection method is introduced to estimate missing data and improve the image quality.
Metal artifacts arise in CT images when X-rays traverse the high attenuating objects such as metal bodies. Portions of projection data become unavailable. In this paper, we present an Euler's elastica and curvature based sinogram inpainting (EECSI) algorithm for metal artifacts reduction, where "inpainting" is a synonym for "image interpolation". In EECSI, the unavailable data are regarded as occlusion and can be inpainted inside the inpainting domain based on elastica interpolants. Numerical simulations demonstrate that, compared to conventional interpolation methods, the algorithm proposed connects the unavailable projection region more smoothly and accurately, thus better reduces metal artifacts and more accurately reveals cross section structures, especially in the immediate neighborhood of the metallic objects.
In this paper we present a backprojection filtered type (BPF-type) reconstruction algorithm for cone-beam circular scans based on Zou and Pan's work. The algorithm could use all the projection data passing through the PI-line segments in 2π scanning range. Because all the projection data in 2π is used, the algorithm has a good quality for practical noisy projection data. The algorithm is implemented using numerical and practical experiments. The practical experiments were done on our X-ray CT system with a flat-panel detector. We also compare the results with FDK reconstructions. From the experimental results, we deem that the BPF algorithm could satisfy the requirement of the X-ray CT inspection.
T-FDK algorithm is an FDK-type cone beam CT reconstruction algorithm. Like other 3D reconstruction algorithms, T-FDK is time consuming because of the large amount of data processing involved. One solution to this problem is utilizing PC graphics boards (GPU) for acceleration. The recent dramatic evolution of GPU makes this method come to the practical track. In this paper, we use a new floating point GPU to speedup the 3D T-FDK algorithm that is different from original FDK method in structure. Because floating point pipelines are slower than hardwired 8-bit texture mapping facilities but are more precise numerically, we balance the reconstruction speed and quality by using both of them. Using nVIDIA GeForce 6800 GT, our GPU accelerated T-FDK method gives a speed 27.612 times faster than a software implementation.
Image segmentation is a classical and challenging problem in image processing and computer vision. Most of the segmentation algorithms, however, do not consider overlapped objects. Due to the special characteristics of X-ray imaging, the overlapping of objects is very commonly seen in X-ray images and needs to be carefully dealt with. In this paper, we propose a novel energy functional to solve this problem. The Euler-Lagrange equation is derived and the segmentation is converted to a front propagating problem that can be efficiently solved by level set methods. We noticed that the proposed energy functional has no unique extremum and the solution relies on the initialization. Thus, an initialization method is proposed to get satisfying results. The experiment on real data validated our proposed method.
Cupping artifact is one of the most serious problems in a middle-low energy X-ray Flat panel detector (FPD)-based cone beam CT system. Both beam hardening effects and scatter could induce cupping artifacts in reconstructions and degrade image quality. In this paper, a two-step cupping-correction method is proposed to eliminate cupping: 1) scatter removal; 2) beam hardening correction. By experimental measurement using Beam Stop Array (BSA), the X-ray scatter distribution of a specific object is estimated in the projection image. After interpolation and subtraction, the primary intensity of the projection image is computed. The scatter distribution can also be obtained using convolution with a low-pass filter as kernel. The linearization is used as beam hardening correction method for one-material object. For two-material cylindrical objects, a new approach without iteration involved is present. There are three processes in this approach. Firstly, correct raw projections by the mapping function of the outer material. Secondly, reconstruct the cross-section image from the modified projections. Finally, scale the image by a simple weighting function. After scatter removal and beam hardening correction, the cupping artifacts are well removed, and the contrast of the reconstructed image is remarkably improved.
Optical Character Recognition (OCR) is a classical research field and has become one of most successful applications in the area of pattern recognition. Feature extraction is a key step in the process of OCR. This paper presents three algorithms for feature extraction based on binary images: the Lattice with Distance Transform (DTL), Stroke Density (SD) and Co-occurrence Matrix (CM). DTL algorithm improves the robustness of the lattice feature by using distance transform to increase the distance of the foreground and background and thus
reduce the influence from the boundary of strokes. SD and CM algorithms extract robust stroke features base on the fact that human recognize characters according to strokes, including length and orientation. SD reflects the quantized stroke information including the length and the orientation. CM reflects the length and orientation of a contour. SD and CM together sufficiently describe strokes. Since these three groups of feature vectors complement each other in expressing characters, we integrate them and adopt a hierarchical algorithm to achieve optimal performance. Our methods are tested on the USPS (United States Postal Service) database and the Vehicle License Plate Number Pictures Database (VLNPD). Experimental results shows that the methods gain high recognition rate and cost reasonable average running time. Also, based on similar condition, we compared our results to the box method proposed by Hannmandlu . Our methods demonstrated better performance in efficiency.
An efficient noise treatment scheme has been developed to achieve low-dose CT diagnosis based on currently available CT hardware and image reconstruction technologies. The scheme proposed includes two main parts: filtering in sinogram domain and smoothing in image domain. The acquired projection sinograms were first treated by our previously proposed Karhunen-Loeve (K-L) domain penalized weighted least-square (PWLS) filtering, which fully utilizes the prior statistical noise property and three-dimensional (3D) spatial information for an accurate restoration of the low-dose projections. To treat the streak artifacts due to photon starvation, we also incorporated an adaptive filtering into our PWLS framework, which selectively smoothes those channels contributing most to the streak artifacts. After the sinogram filtering, the image was reconstructed by the conventional filtered backprojection (FBP) method. The image is assumed as piecewise regions each has a unique texture. Therefore, an edge-preserving smoothing (EPS) with locally-adaptive parameters to the noise variation was applied for further noise reduction in image domain. Experimental phantom projections acquired by a GE spiral computed tomography (CT) scanner under 10 mAs tube current were used to evaluate the proposed smoothing scheme. The reconstructed imaged demonstrated that the smoothing scheme with appropriate control parameters provides a significant improvement on noise suppression without sacrificing the spatial resolution.