Prior Information Induced Sparse-view CT Image Reconstruction

X-ray computed tomography (CT) has become an excellent example for medical imaging with high temporal resolution, spatial resolution and contrast resolution. It has been widely used in the clinical diagnosis and treatment. But the excessive X-ray radiation may induce cancer, leukemia, or some other hereditary disease, so the safety and security of CT imaging has become one focus for the industry. How to get the best quality of CT images with the minimum cost and minimum X-ray dose has become one of the most important issues.Various techniques to reduce radiation dose in CT scans have been investigated. The most typical one is the low X-ray tube milliampere seconds (mAs) technique by reducing the tube current and exposure time. In addition, down sampling the viewing angle is another important way to reduce the radiation dose in CT scans. Meanwhile, the associated reconstructed image usually suffers from serious noise and artifacts, which has negative influence for clinical diagnosis. By reducing the number of incident photons, low mAs technique can achieve radiation dose reduction. The projection data is characterized by severe noise, followed by the significantly degraded image quality. Recently, the image reconstruction methods for the low-mAs technique can be divided into three strategies. The first one is implementing filtering in image domain and reducing noise and artifacts directly. This strategy belongs to image re-processing technique. The second one is building projection restoration model based on the noise distribution characteristics. Implementing projection restoration and then reconstructing with filtered back-projection (FBP) method from the restored projection data. The third one is reconstruction in image domain directly based on the projection noise distribution. By reducing the number of projection per rotation can achieve sparse-view technique. However, due to insufficient sampling of sparse-view CT measurements, conventional FBP approach cannot yield high diagnostic image quality. Until now, various algorithms for the sparse-view image reconstruction have been investigated, which mainly includes projection interpolation or spatial transformation algorithms and iterative image reconstruction algorithms based on compressed sensing theory. Due to the lack of original projection data, simple interpolation or spatial transformation method cannot obtain satisfied results, which can unavoidably leads to the loss of the original structures and local details. The iterative reconstruction algorithm can overcome the inherent physical limitations of CT by modeling the imaging geometry, X-ray spectrum, beam hardening, scattering, and noise, etc. Furthermore, iterative reconstruction algorithms that based on photon statistics constructing an accurate noise model, has a better performance in reduce image noise. Therefore, the iterative reconstruction algorithm can further improve the image spatial resolution and reduce the artllifacts.In recent years, researches about sparse-view image reconstruction are very hot, which is mainly around the compressed sensing algorithm. The CS theory provides a new theoretical basis dealing with the problem of incomplete projection data reconstruction. Essentially, the problem of incomplete CT projection data reconstruction comes from the problem of inadequate sampling in Fourier space. The sampling frequency is much lower than the Nyquist sampling frequency. The CS theory proposed by Candes provided the method to overcome this problem. The CS theory shows that high-quality signals and images can be reconstructed from far less data/measurements than what is usually considered necessary according to the Nyquist sampling theory. Because of the sparse projection data, the reconstruction image contains serious streak artifacts. Based on that, Sidky proposed the Total Variation method, which fulfills accurate reconstruction from far less data when the data is sparse. The TV minimization method consists of two phases:POCS and gradient descent. The POCS phase constrains the consistency and the nonnegative of the image. The gradient descent phase is used to solve the TV minimization to obtain a new image estimate. Hence, the TV minimization based adaptive steepest-descent POCS algorithm (ASD-POCS) as an extension of the original one were proposed recently for sparse-view CT image reconstructionThere are two general categories of iterative image reconstruction algorithms based on CS theory, the algebraic reconstruction technique (ART) based CS and the statistical iterative reconstruction (SIR) based CS. ART method is based upon the theory that the secondary projection in the topographic measurements can be regarded as a set of projections of the cross-section of some unknown target function. Then the method combined TV with POCS was proposed, which achieved successful reconstruction results. However, the major drawback of the ART method is that the statistical properties of CT measurement cannot be well considered. Compared with ART method, iterative reconstruction algorithms that based on photon statistics constructing an accurate noise model has a better performance in reduce image noise. Generally, accurate modeling of projection data measurements is the foundation of high quality reconstruction, and the introduction of prior knowledge to avoid solving ambiguity and high precision image reconstruction has very important significance.X-ray CT technologies have been widely explored for specific applications in clinic including perfusion imaging,4D-CT imaging, image-guided intervention and radiotherapy, et al. Under these situations, repeated tomographic acquisitions are often prescribed. For instance, except the planning CT, in daily CBCT examinations for target localization in image-guided radiation therapy (IGRT), repeated scans have become routine procedures. In this case, the cumulative radiation dose still significantly increase as comparison with the conventional CT scans, which has raised major concerns in patients. With regard to the repeated CT scans, a previously scanned high-quality diagnostic CT image volume usually contains same anatomical information as the current scan except for some anatomical changes due to internal motion or patient weight change. In other words, there exists redundant information among the repeated scan CT images. However, the patient position is frequently changed from one scan to another within the time-series data acquirement. Additionally, vessels and tissues would change their attenuation properties after the intravascular contrast agent mixed with blood. In these cases, the mismatched areas in previously scanned image would inevitably influence the final reconstruction results. So, how to use previously scanned information to improve the final image quality in sparse-view image reconstruction, and reduce the influence of the misalignment resulting from motion become the focus in this research.Through combining low-mAs protocol and sparse-view protocol can realize low-dose CT imaging, which is a new but important topic in imaging techniques. For the combined low-mAs and sparse-view problem, low-mAs scan can lead to noisy projection, which can cause the degradation of sparse-view reconstruction image. So, how to fulfill a preprocessing step for noisy sinogram and get both noise reduced and artifacts suppressed images become the focus in this research.On the whole, the main works of this paper can be summarized as follows:(1) To introduce the prior image induced TV method for sparse-view image reconstruction. The sparse-view image quality can be improved by inducing previously scanned normal-dose image. To overcome the disadvantage of the locally designed prior term with only the different values of the neighboring pixels as the constraint, in this paper, we propose to utilize the nonlocal criteria to search the patched based similarity between different pixels. In this way, the prior information can be introduced efficiently. This algorithm was conducted in the frame of penalized weighted least-square (PWLS) approach, combined with the total variation (TV) method. Based on our previous work, an alternating minimization scheme was used. The reconstruction image and the weights quantifying the similarity between the sparse-view image and the normal-dose image were updated in the iteration. Qualitative and quantitative evaluations or CT image reconstruction from sparse-view projection data were carried out on two digital phantoms and an anthropomorphic torso phantom in terms of image visualization, resolution-noise tradeoff, and noise reduction metrics (peak signal-to noise ratio, mean per cent squared error and mean per cent absolute error). Experimental results show that the present approach can relieve the requirement of misalignment reduction of the other approach and can achieve significant gains over the existing similar methods in terms of noise and artifacts suppression and edges information preservation.(2) To introduce the sinogram restoration induced ultra-low-dose CT image reconstruction. The proposed method contains sinogram restoration preprocessing step and image iterative reconstruction step. Specifically, the sinogram data acquired with a combined low-mAs and sparse-view protocol is first restored by using a PWLS based sinogram restoration method. Then, the restored sinogram data is hereafter used to reconstruct image by using a PWLS based total variation (PWLS-TV) method. Qualitative and quantitative evaluations or CT image reconstruction from sparse-view projection data were carried out on digital NCAT phantom and real data. To validate and evaluate the performance of the proposed method, projection data with different levels of mAs and different number of views were simulated. For quantitative evaluation of the reconstruction accuracy, the relative root mean squared error (RRMSE) and per cent absolute error (MPAE) metrics is calculated. Experiment results show that the proposed method can achieve successful image reconstruction with the restored sinogram, which can lead to the reduction of radiation dose. It is worth to note that The PWLS sinogram restoration as a preprocessing step is useful for the ultra-low-dose CT image reconstruction with a combined low-mAs and sparse-view protocol.