Artificial intelligence in image reconstruction: The change is here

Innovations in CT have been impressive among imaging and medical technologies in both the hardware and software domain. The range and speed of CT scanning improved from the introduction of multidetector-row CT scanners with wide-array detectors and faster gantry rotation speeds. To tackle concerns over rising radiation doses from its increasing use and to improve image quality, CT reconstruction techniques evolved from filtered back projection to commercial release of iterative reconstruction techniques, and recently, of deep learning (DL)-based image reconstruction. These newer reconstruction techniques enable improved or retained image quality versus filtered back projection at lower radiation doses. DL can aid in image reconstruction with training data without total reliance on the physical model of the imaging process, unique artifacts of PCD-CT due to charge sharing, K-escape, fluorescence x-ray emission, and pulse pileups can be handled in the data-driven fashion. With sufficiently reconstructed images, a well-designed network can be trained to upgrade image quality over a practical/clinical threshold or define new/killer applications. Besides, the much smaller detector pixel for PCD-CT can lead to huge computational costs with traditional model-based iterative reconstruction methods whereas deep networks can be much faster with training and validation. In this review, we present techniques, applications, uses, and limitations of deep learning-based image reconstruction methods in CT.     

System Matrix Analysis for Computed Tomography Imaging

In practical applications of computed tomography imaging (CT), it is often the case that the set of projection data is incomplete owing to the physical conditions of the data acquisition process. On the other hand, the high radiation dose imposed on patients is also undesired. These issues demand that high quality CT images can be reconstructed from limited projection data. For this reason, iterative methods of image reconstruction have become a topic of increased research interest. Several algorithms have been proposed for few-view CT. We consider that the accurate solution of the reconstruction problem also depends on the system matrix that simulates the scanning process. In this work, we analyze the application of the Siddon method to generate elements of the matrix and we present results based on real projection data.     

An Effective CUDA Parallelization of Projection in Iterative Tomography Reconstruction

Projection and back-projection are the most computationally intensive parts in Computed Tomography (CT) reconstruction, and are essential to acceleration of CT reconstruction algorithms. Compared to back-projection, parallelization efficiency in projection is highly limited by racing condition and thread unsynchronization. In this paper, a strategy of Fixed Sampling Number Projection (FSNP) is proposed to ensure the operation synchronization in the ray-driven projection with Graphical Processing Unit (GPU). Texture fetching is also used utilized to further accelerate the interpolations in both projection and back-projection. We validate the performance of this FSNP approach using both simulated and real cone-beam CT data. Experimental results show that compare to the conventional approach, the proposed FSNP method together with texture fetching is 10~16 times faster than the conventional approach based on global memory, and thus leads to more efficient iterative algorithm in CT reconstruction.     

Reduction of Metal Artifact in Single Photon-Counting Computed Tomography by Spectral-Driven Iterative Reconstruction Technique

PurposeThe exciting prospect of Spectral CT (SCT) using photon-counting detectors (PCD) will lead to new techniques in computed tomography (CT) that take advantage of the additional spectral information provided. We introduce a method to reduce metal artifact in X-ray tomography by incorporating knowledge obtained from SCT into a statistical iterative reconstruction scheme. We call our method Spectral-driven Iterative Reconstruction (SPIR).MethodThe proposed algorithm consists of two main components: material decomposition and penalized maximum likelihood iterative reconstruction. In this study, the spectral data acquisitions with an energy-resolving PCD were simulated using a Monte-Carlo simulator based on EGSnrc C++ class library. A jaw phantom with a dental implant made of gold was used as an object in this study. A total of three dental implant shapes were simulated separately to test the influence of prior knowledge on the overall performance of the algorithm. The generated projection data was first decomposed into three basis functions: photoelectric absorption, Compton scattering and attenuation of gold. A pseudo-monochromatic sinogram was calculated and used as input in the reconstruction, while the spatial information of the gold implant was used as a prior. The results from the algorithm were assessed and benchmarked with state-of-the-art reconstruction methods.ResultsDecomposition results illustrate that gold implant of any shape can be distinguished from other components of the phantom. Additionally, the result from the penalized maximum likelihood iterative reconstruction shows that artifacts are significantly reduced in SPIR reconstructed slices in comparison to other known techniques, while at the same time details around the implant are preserved. Quantitatively, the SPIR algorithm best reflects the true attenuation value in comparison to other algorithms.ConclusionIt is demonstrated that the combination of the additional information from Spectral CT and statistical reconstruction can significantly improve image quality, especially streaking artifacts caused by the presence of materials with high atomic numbers.     

Quantitative study on exact reconstruction sampling condition by verifying solution uniqueness in limited-view CT

In limited-view computed tomography reconstruction, iterative image reconstruction with sparsity-exploiting methods, such as total variation (TV) minimization, inspired by compressive sensing, potentially claims large reductions in sampling requirements. However, a quantitative notion of this claim is non-trivial because of the ill-defined reduction in sampling achieved by the sparsity-exploiting method. In this paper, exact reconstruction sampling condition for limited-view problem is studied by verifying the uniqueness of solution in TV minimization model. Uniqueness is tested by solving a convex optimization problem derived from the sufficient and necessary condition of solution uniqueness. Through this method, the sufficient sampling number of exact reconstruction is quantified for any fixed phantom and settled geometrical parameter in the limited-view problem. This paper provides a reference to quantify the sampling condition. Three phantoms are tested to study the sampling condition of limited view exact reconstruction in this paper. The experiment results show the quantified sampling number and indicate that an object would be accurately reconstructed as the scanning range becomes narrower by increasing sampling number. The increased samplings compensate for the deficiency of the projection angle. However, the lower bound of the scanning range corresponding to three different phantoms are presented, in which an exact reconstruction cannot be obtained once the projection angular is narrowed to this extent no matter how to increase sampling.     

Iterative Image Reconstruction for Sparse-View CT Using Normal-Dose Image Induced Total Variation Prior

X-ray computed tomography (CT) iterative image reconstruction from sparse-view projection data has been an important research topic for radiation reduction in clinic. In this paper, to relieve the requirement of misalignment reduction operation of the prior image constrained compressed sensing (PICCS) approach introduced by Chen et al, we present an iterative image reconstruction approach for sparse-view CT using a normal-dose image induced total variation (ndiTV) prior. The associative objective function of the present approach is constructed under the penalized weighed least-square (PWLS) criteria, which contains two terms, i.e., the weighted least-square (WLS) fidelity and the ndiTV prior, and is referred to as “PWLS-ndiTV”. Specifically, the WLS fidelity term is built based on an accurate relationship between the variance and mean of projection data in the presence of electronic background noise. The ndiTV prior term is designed to reduce the influence of the misalignment between the desired- and prior- image by using a normal-dose image induced non-local means (ndiNLM) filter. Subsequently, a modified steepest descent algorithm is adopted to minimize the associative objective function. Experimental results on two different digital phantoms and an anthropomorphic torso phantom show that the present PWLS-ndiTV approach for sparse-view CT image reconstruction can achieve noticeable gains over the existing similar approaches in terms of noise reduction, resolution-noise tradeoff, and low-contrast object detection.     

Statistical learning methods as a preprocessing step for survival analysis: evaluation of concept using lung cancer data

Background

Despite its superb lateral resolution, flat-panel-detector (FPD) based tomosynthesis suffers from low contrast and inter-plane artifacts caused by incomplete cancellation of the projection components stemming from outside the focal plane. The incomplete cancellation of the projection components, mostly due to the limited scan angle in the conventional tomosynthesis scan geometry, often makes the image contrast too low to differentiate the malignant tissues from the background tissues with confidence.

Methods

In this paper, we propose a new method to suppress the inter-plane artifacts in FPD-based tomosynthesis. If 3D whole volume CT images are available before the tomosynthesis scan, the CT image data can be incorporated into the tomosynthesis image reconstruction to suppress the inter-plane artifacts, hence, improving the image contrast. In the proposed technique, the projection components stemming from outside the region-of-interest (ROI) are subtracted from the measured tomosynthesis projection data to suppress the inter-plane artifacts. The projection components stemming from outside the ROI are calculated from the 3D whole volume CT images which usually have lower lateral resolution than the tomosynthesis images. The tomosynthesis images are reconstructed from the subtracted projection data which account for the x-ray attenuation through the ROI. After verifying the proposed method by simulation, we have performed both CT scan and tomosynthesis scan on a phantom and a sacrificed rat using a FPD-based micro-CT.

Results

We have measured contrast-to-noise ratio (CNR) from the tomosynthesis images which is an indicator of the residual inter-plane artifacts on the focal-plane image. In both cases of the simulation and experimental imaging studies of the contrast evaluating phantom, CNRs have been significantly improved by the proposed method. In the rat imaging also, we have observed better visual contrast from the tomosynthesis images reconstructed by the proposed method.

Conclusions

The proposed tomosynthesis technique can improve image contrast with aids of 3D whole volume CT images. Even though local tomosynthesis needs extra 3D CT scanning, it may find clinical applications in special situations in which extra 3D CT scan is already available or allowed.     

Iterative image reconstruction using modified non-local means filtering for limited-angle computed tomography

PurposeLimited-angle CT imaging is an effective technique to reduce radiation. However, existing image reconstruction methods can effectively reduce streak artifacts but fail to suppress those artifacts around edges due to incomplete projection data. Thus, a modified NLM (mNLM) based reconstruction method is proposed.MethodsSince the artifacts around edges mainly exist in local position, it is possible to restore the true pixels in artifacts using pixels located in artifacts-free regions. In each iteration, mNLM is performed on image reconstructed by ART followed by positivity constraint. To solve the problem caused by ART-mNLM that there is undesirable information that may appear in the image, ART-TV is then utilized in the following iterative process after ART-mNLM iterates for a number of iterations. The proposed algorithm is named as ART-mNLM/TV.ResultsSimulation experiments are performed to validate the feasibility of algorithm. When the scanning range is [0, 150°], our algorithm outperforms the ART-NLM and ART-TV with more than 40% and 29% improvement in terms of SNR and with more than 58% and 49% reduction in terms of MAE. Consistently, reconstructed images from real projection data also demonstrate the effectiveness of presented algorithm.ConclusionThis paper uses mNLM which benefits from redundancy of information across the whole image, to recover the true value of pixels in artifacts region by utilizing pixels from artifact-free regions, and artifacts around the edges can be mitigated effectively. Experiments show that the proposed ART-mNLM/TV is able to achieve better performances compared to traditional methods.     

CT iterative reconstruction in image space: A phantom study

Although iterative reconstruction is widely applied in SPECT/PET, its introduction in clinical CT is quite recent, in the past the demand for extensive computer power and long image reconstruction times have stopped the diffusion of this technique. Recently Iterative Reconstruction in Image Space (IRIS) has been introduced on Siemens top CT scanners. This recon method works on image data area, reducing the time-consuming loops on raw data and noise removal is obtained in subsequent iterative steps with a smoothing process. We evaluated image noise, low contrast resolution, CT number linearity and accuracy, transverse and z-axis spatial resolution using some dedicated phantoms in single, dual source and cardiac mode. We reconstructed images with a traditional filtered back-projection algorithm and with IRIS. The iterative procedure preserves spatial resolution, CT number accuracy and linearity moreover decreases image noise. These preliminary results support the idea that dose reduction with preserved image quality is possible with IRIS, even if studies on patients are necessary to confirm these data.     

A Novel Iterative CT Reconstruction Approach Based on FBP Algorithm

The Filtered Back-Projection (FBP) algorithm and its modified versions are the most important techniques for CT (Computerized tomography) reconstruction, however, it may produce aliasing degradation in the reconstructed images due to projection discretization. The general iterative reconstruction (IR) algorithms suffer from their heavy calculation burden and other drawbacks. In this paper, an iterative FBP approach is proposed to reduce the aliasing degradation. In the approach, the image reconstructed by FBP algorithm is treated as the intermediate image and projected along the original projection directions to produce the reprojection data. The difference between the original and reprojection data is filtered by a special digital filter, and then is reconstructed by FBP to produce a correction term. The correction term is added to the intermediate image to update it. This procedure can be performed iteratively to improve the reconstruction performance gradually until certain stopping criterion is satisfied. Some simulations and tests on real data show the proposed approach is better than FBP algorithm or some IR algorithms in term of some general image criteria. The calculation burden is several times that of FBP, which is much less than that of general IR algorithms and acceptable in the most situations. Therefore, the proposed algorithm has the potential applications in practical CT systems.     

Sparse-view CT reconstruction based on multi-level wavelet convolution neural network

Sparse-view computed tomography (CT) is a recent approach to reducing the radiation dose in patients and speeding up the data acquisition. Consequently, sparse-view CT has been of particular interest among researchers within the CT community. Advanced reconstruction algorithms for sparse-view CT, such as iterative algorithms with total-variation (TV), have been studied along with the problem of increasing computational burden and the blurring of artifacts in the reconstructed images. Studies on deep-learning-based approaches applying U-NET have recently achieved remarkable outcomes in various domains including low-dose CT. In this study, we propose a new method for sparse-view CT reconstruction based on a multi-level wavelet convolutional neural network (MWCNN). First, a filtered backprojection (FBP) was used to reconstruct a sparsely sampled sinogram from 60, 120, and 180 projections. Subsequently, the sparse-view data obtained from FBP were fed to a deep-learning network, i.e., the MWCNN. Our network architecture combines a wavelet transform and modified U-NET without pooling. By replacing the pooling function with the wavelet transform, the receptive field is enlarged to improve the performance. We qualitatively and quantitatively evaluated the interpolation, iterative TV method, and standard U-NET in terms of a reduction in the streaking artifacts and a preservation of the anatomical structures. When compared with other methods, the proposed method showed the highest performance based on various evaluation parameters such as the structural similarity, root mean square error, and resolution. These results indicate that the MWCNN possesses a powerful potential for achieving a sparse-view CT reconstruction.     

An iterative deconvolution algorithm for image recovery in clinical CT: A phantom study

PurposeTo study the feasibility of using an iterative reconstruction algorithm to improve previously reconstructed CT images which are judged to be non-diagnostic on clinical review. A novel rapidly converging, iterative algorithm (RSEMD) to reduce noise as compared with standard filtered back-projection algorithm has been developed.Materials and methodsThe RSEMD method was tested on in-silico, Catphan^®500, and anthropomorphic 4D XCAT phantoms. The method was applied to noisy CT images previously reconstructed with FBP to determine improvements in SNR and CNR. To test the potential improvement in clinically relevant CT images, 4D XCAT phantom images were used to simulate a small, low contrast lesion placed in the liver.ResultsIn all of the phantom studies the images proved to have higher resolution and lower noise as compared with images reconstructed by conventional FBP. In general, the values of SNR and CNR reached a plateau at around 20 iterations with an improvement factor of about 1.5 for in noisy CT images. Improvements in lesion conspicuity after the application of RSEMD have also been demonstrated. The results obtained with the RSEMD method are in agreement with other iterative algorithms employed either in image space or with hybrid reconstruction algorithms.ConclusionsIn this proof of concept work, a rapidly converging, iterative deconvolution algorithm with a novel resolution subsets-based approach that operates on DICOM CT images has been demonstrated. The RSEMD method can be applied to sub-optimal routine-dose clinical CT images to improve image quality to potentially diagnostically acceptable levels.     

Simulation of scanner- and patient-specific low-dose CT imaging from existing CT images

PurposeSimulating low-dose Computed Tomography (CT) facilitates in-silico studies into the required dose for a diagnostic task. Conventionally, low-dose CT images are created by adding noise to the projection data. However, in practice the raw data is often simply not available. This paper presents a new method for simulating patient-specific, low-dose CT images without the need of the original projection data.MethodsThe low-dose CT simulation method included the following: (1) computation of a virtual sinogram from a high dose CT image through a radon transform; (2) simulation of a ‘reduced’-dose sinogram with appropriate amounts of noise; (3) subtraction of the high-dose virtual sinogram from the reduced-dose sinogram; (4) reconstruction of a noise volume via filtered back-projection; (5) addition of the noise image to the original high-dose image. The required scanner-specific parameters, such as the apodization window, bowtie filter, the X-ray tube output parameter (reflecting the photon flux) and the detector read-out noise, were retrieved from calibration images of a water cylinder. The low-dose simulation method was evaluated by comparing the noise characteristics in simulated images with experimentally acquired data.ResultsThe models used to recover the scanner-specific parameters fitted accurately to the calibration data, and the values of the parameters were comparable to values reported in literature. Finally, the simulated low-dose images accurately reproduced the noise characteristics in experimentally acquired low-dose-volumes.ConclusionThe developed methods truthfully simulate low-dose CT imaging for a specific scanner and reconstruction using filtered backprojection. The scanner-specific parameters can be estimated from calibration data.     

Enhancement of soft-tissue contrast in cone-beam CT using an anti-scatter grid with a sparse sampling approach

PurposeAnti-scatter grids suppress the scatter substantially thus improving image contrast in radiography. However, its active use in cone-beam CT for the purpose of improving contrast-to-noise ratio (CNR) has not been successful mainly due to the increased noise related to Poisson statistics of photons. This paper proposes a sparse-view scanning approach to address the above issue.MethodCompared to the conventional cone-beam CT imaging framework, the proposed method reduces the number of projections and increases exposure in each projection to enhance image quality without an additional cost of radiation dose to patients. For image reconstruction from sparse-view data, an adaptive-steepest-descent projection-onto-convex-sets (ASD POCS) algorithm regularized by total-variation (TV) minimization was adopted. Contrast and CNR with various scattering conditions were evaluated in projection domain by a simulation study using GATE. Then we evaluated contrast, resolution, and image uniformity in CT image domain with Catphan phantom. A head phantom with soft-tissue structures was also employed for demonstrating a realistic application. A virtual grid-based estimation and reduction of scatter has also been implemented for comparison with the real anti-scatter grid.ResultsIn the projection domain evaluation, contrast and CNR enhancement was observed when using an anti-scatter grid compared to the virtual grid. In the CT image domain, the proposed method produced substantially higher contrast and CNR of the low-contrast structures with much improved image uniformity.ConclusionWe have shown that the proposed method can provide high-quality CBCT images particularly with an increased contrast of soft-tissue at a neutral dose for image-guidance.     

Iterative reconstruction methods in X-ray CT

Iterative reconstruction (IR) methods have recently re-emerged in transmission x-ray computed tomography (CT). They were successfully used in the early years of CT, but given up when the amount of measured data increased because of the higher computational demands of IR compared to analytical methods. The availability of large computational capacities in normal workstations and the ongoing efforts towards lower doses in CT have changed the situation; IR has become a hot topic for all major vendors of clinical CT systems in the past 5 years.This review strives to provide information on IR methods and aims at interested physicists and physicians already active in the field of CT. We give an overview on the terminology used and an introduction to the most important algorithmic concepts including references for further reading. As a practical example, details on a model-based iterative reconstruction algorithm implemented on a modern graphics adapter (GPU) are presented, followed by application examples for several dedicated CT scanners in order to demonstrate the performance and potential of iterative reconstruction methods. Finally, some general thoughts regarding the advantages and disadvantages of IR methods as well as open points for research in this field are discussed.     

Single-particle reconstruction using L(2)-gradient flow

In this paper, we present an iterative algorithm for reconstructing a three-dimensional density function from a set of two dimensional electron microscopy images. By minimizing an energy functional consisting of a fidelity term and a regularization term, an L²-gradient flow is derived. The flow is integrated by a finite element method in the spatial direction and an explicit Euler scheme in the temporal direction. Our method compares favorably with those of the weighted back projection, Fourier method, algebraic reconstruction technique and simultaneous iterative reconstruction technique.     

Accelerated Optical Projection Tomography Applied to In Vivo Imaging of Zebrafish

Optical projection tomography (OPT) provides a non-invasive 3-D imaging modality that can be applied to longitudinal studies of live disease models, including in zebrafish. Current limitations include the requirement of a minimum number of angular projections for reconstruction of reasonable OPT images using filtered back projection (FBP), which is typically several hundred, leading to acquisition times of several minutes. It is highly desirable to decrease the number of required angular projections to decrease both the total acquisition time and the light dose to the sample. This is particularly important to enable longitudinal studies, which involve measurements of the same fish at different time points. In this work, we demonstrate that the use of an iterative algorithm to reconstruct sparsely sampled OPT data sets can provide useful 3-D images with 50 or fewer projections, thereby significantly decreasing the minimum acquisition time and light dose while maintaining image quality. A transgenic zebrafish embryo with fluorescent labelling of the vasculature was imaged to acquire densely sampled (800 projections) and under-sampled data sets of transmitted and fluorescence projection images. The under-sampled OPT data sets were reconstructed using an iterative total variation-based image reconstruction algorithm and compared against FBP reconstructions of the densely sampled data sets. To illustrate the potential for quantitative analysis following rapid OPT data acquisition, a Hessian-based method was applied to automatically segment the reconstructed images to select the vasculature network. Results showed that 3-D images of the zebrafish embryo and its vasculature of sufficient visual quality for quantitative analysis can be reconstructed using the iterative algorithm from only 32 projections—achieving up to 28 times improvement in imaging speed and leading to total acquisition times of a few seconds.     

基于CT薄层扫描的骨盆数字化三维模型的分色构建

目的：探讨利用CT原始数据集对骨盆进行数字化三维分色构建的方法及意义。方法：选择1例因宫颈癌行盆腔CT薄层扫描患者的Dicom3．0原始二维断层数据集,利用Mimics10．01软件行骨盆三维分色重建。结果：构建的数字化三维分色模型形态规则、清晰逼真、立体感强、解剖清晰,不仅可以对构成骨盆的髂骨、骶骨及尾骨进行单独的三维分色显示,而且可以进行任意角度、距离的融合分离显示,更有利于对骨盆进行精细地全面立体观察分析。结论：基于CT薄层扫描数据集构建骨盆三维分色模型的方法简单、可行,是指导临床及教学的好工具。     

A new high‐resolution computed tomography (CT) segmentation method for trabecular bone architectural analysis

In the last decade, high‐resolution computed tomography (CT) and microcomputed tomography (micro‐CT) have been increasingly used in anthropological studies and as a complement to traditional histological techniques. This is due in large part to the ability of CT techniques to nondestructively extract three‐dimensional representations of bone structures. Despite prior studies employing CT techniques, no completely reliable method of bone segmentation has been established. Accurate preprocessing of digital data is crucial for measurement accuracy, especially when subtle structures such as trabecular bone are investigated. The research presented here is a new, reproducible, accurate, and fully automated computerized segmentation method for high‐resolution CT datasets of fossil and recent cancellous bone: the Ray Casting Algorithm (RCA). We compare this technique with commonly used methods of image thresholding (i.e., the half‐maximum height protocol and the automatic, adaptive iterative thresholding procedure). While the quality of the input images is crucial for conventional image segmentation, the RCA method is robust regarding the signal to noise ratio, beam hardening, ring artifacts, and blurriness. Tests with data of extant and fossil material demonstrate the superior quality of RCA compared with conventional thresholding procedures, and emphasize the need for careful consideration of optimal CT scanning parameters. Am J Phys Anthropol 2009. © 2009 Wiley‐Liss, Inc.