Iterative Algebraic Reconstruction Technique Research Articles

Mitigation of motion-induced artifacts in cone beam computed tomography using deep convolutional neural networks.

Cone beam computed tomography (CBCT) is often employed on radiation therapy treatment devices (linear accelerators) used in image-guided radiation therapy (IGRT). For each treatment session, it is necessary to obtain the image of the day in order to accurately position the patient and to enable adaptive treatment capabilities including auto-segmentation and dose calculation. Reconstructed CBCT images often suffer from artifacts, in particular those induced by patient motion. Deep-learning based approaches promise ways to mitigate such artifacts. We propose a novel deep-learning based approach with the goal to reduce motion induced artifacts in CBCT images and improve image quality. It is based on supervised learning and includes neural network architectures employed as pre- and/or post-processing steps during CBCT reconstruction. Our approach is based on deep convolutional neural networks which complement the standard CBCT reconstruction, which is performed either with the analytical Feldkamp-Davis-Kress (FDK) method, or with an iterative algebraic reconstruction technique (SART-TV). The neural networks, which are based on refined U-net architectures, are trained end-to-end in a supervised learning setup. Labeled training data are obtained by means of a motion simulation, which uses the two extreme phases of 4D CT scans, their deformation vector fields, as well as time-dependent amplitude signals as input. The trained networks are validated against ground truth using quantitative metrics, as well as by using real patient CBCT scans for a qualitative evaluation by clinicalexperts. The presented novel approach is able to generalize to unseen data and yields significant reductions in motion induced artifacts as well as improvements in image quality compared with existing state-of-the-art CBCT reconstruction algorithms (up to +6.3 dB and +0.19 improvements in peak signal-to-noise ratio, PSNR, and structural similarity index measure, SSIM, respectively), as evidenced by validation with an unseen test dataset, and confirmed by a clinical evaluation on real patient scans (up to 74% preference for motion artifact reduction over standard reconstruction). For the first time, it is demonstrated, also by means of clinical evaluation, that inserting deep neural networks as pre- and post-processing plugins in the existing 3D CBCT reconstruction and trained end-to-end yield significant improvements in image quality and reduction of motionartifacts.

Robust extension of light fields with probable 3D distribution based on iterative scene estimation from multi-focus images

In this study, we propose an iterative approach to perform pseudo-scene reconstruction based on multi-focus images for estimating the 4D light fields of the scene robustly. An analogous scene-independent linear filtering approach has been proposed to efficiently achieve free viewpoint image reconstruction and scene refocusing with virtual bokeh by solely extracting 3D frequency components. However, it only succeeded in synthesizing limited regions of 4D light fields from multi-focus images, which can lead to awful degradation of reconstructed free viewpoint images outside the range due to the lack of the corresponding frequency components. In this paper, we directly utilize a linear relationship of 3D convolution combining multi-focus images and 3D scene by 3D blurring filters to estimate the 3D distribution itself of the scene instead of deriving the scene-independent linear filters from the relationship. In other words, we perform the scene reconstruction as a simplified convex optimization problem by modeling multi-focus images in multi-layered blurring composed of the scene and a bokeh filter. Algebraic reconstruction techniques of computed tomography, which uses the classical steepest descent method combined with projection onto convex sets (POCS) to ensure the convexity of the problem constraints, are introduced into such 3D scene reconstruction. The new method can be highly anticipated to assist us in extensively estimating robust 4D light fields. Experimental results illustrate that this approach enables good quality 3D image reconstruction with a wider range of 4D light fields. We also discuss the reduction of high computational costs and enormous memory requirements imposed by iterative algebraic reconstruction techniques. The quality of the obtained 4D light fields is comprehensively evaluated using both synthetic and real multi-focus images of practicable image resolution and depth range, additionally to an analysis of the faster processing times when comparing our iterative approach to the performance of the prior linear filters.

Multi view interferometric tomography measurements of convective phenomena in a differentially-heated nanofluid layer

ABSTRACT Three-dimensional reconstructions of thermal field in a differentially heated cylindrical cavity with dilute concentrations of nanofluids as the working medium have been reported. Experiments have been conducted in a specially designed tomography cell filled with dilute concentrations of Al2O3 nanofluids with ethylene glycol dimethacrylate (EGD) as the basefluid. The two horizontal walls of the cell have been differentially heated to apply the temperature potential across the fluid layer. Multiple projection data in the form of interferometric images have been recorded at every 20° view angle by slowly turning the cell over the range of 0 to 360°. The two-dimensional temperature fields retrieved through the quantitative analysis of interferometric images recorded at various view angles have been used to reconstruct the complete three-dimensional temperature field at select horizontal planes of the fluid layer. Iterative multiplicative algebraic reconstruction technique (MART) has been employed for tomography reconstructions. Based on the tomography reconstructions, possible fluid movement in the test cavity and the strong dependence of convective field on varying nanofluid concentrations and cavity temperature differences have been explained. Distribution of local temperature fields near the cavity wall surfaces revealed the phenomena of breakdown of large length scale fluid structures into multiple roll-like patterns with increasing concentration of nanoparticles. The distribution of the local Nusselt number, on select horizontal planes near the top and bottom wall surfaces of the cavity, highlighted the possible role of nanoparticles enhancing the heat transfer rates. An overall energy balance in the cavity was also to be seen based on the reconstructions of the local profiles of Nusselt numbers.

Novel 3-fold metal artifact reduction method for CT images

Metal-induced artifacts are one of the most common streaking artifacts in CT images which are caused by metallic implants within the patient’s body. These streaking artifacts are due to attenuation of X-ray photons passing through these objects, resulting in truncated projections, where gaps are formed in projection data. Metal artifacts can corrupt CT images such that they become challenging to read and interpret, eventually leading to limited diagnostic value. In this paper, a novel 3-Fold metal artifact reduction (MAR) method is proposed, where CT slices which are adjacent to artifact affected are used to correct metal artifact affected CT image. It is a hybrid approach involving the sinogram interpolation correction, projection-profile prediction and iterative algebraic reconstruction technique (ART). The proposed MAR method is compared with recently reported iterative frequency split–normalized (IFS) and L0 – Douglas-Rachford splitting (L0-DRS) MAR approaches using clinical cases of hip prostheses, maintaining the similar experimental environment. Performance comparison by means of 'reference free ground truth' metric revealed greater ability and acceptability of proposed methods. 3-Fold MAR method emerged as most promising technique, with excellent quantitative and qualitative score.

Increasing the Performance of the Iterative Computed Tomography Image Reconstruction Algorithms

Computed tomography (CT) imaging is an important diagnostic tool. CT imaging facilitates the internal rendering of a scanned object by measuring the attenuation of beams of X-ray radiation. CT employs a mathematical technique of image reconstruction; those techniques are classified as; analytical and iterative. The iterative reconstruction (IR) methods have been proven to be superior over the analytical methods, but due to their prolonged reconstruction time, those methods are excluded from routine use in clinical applications. In this paper the reconstruction time of an IR algorithm is minimized through the employment of an adaptive region growing segmentation method that focuses the image reconstruction process on a specified region, thus ignoring unwanted pixels that increase the computation time. This method is tested on the iterative algebraic reconstruction technique (ART) algorithm. Some phantom images are used in this paper to demonstrate the effects of the segmentation process. The simulation results are executed using MATLAB (version R2018b) programming language, and a computer system with the following specifications: CPU core i7 (2.40 GHz) for processing. Simulation results indicate that this method will reduce the reconstruction time of the iterative algorithms, and will enhance the quality of the reconstructed image.

Iterative reconstruction with boundary detection for carbon ion computed tomography

In heavy ion radiation therapy, improving the accuracy in range prediction of the ions inside the patient’s body has become essential. Accurate localization of the Bragg peak provides greater conformity of the tumor while sparing healthy tissues. We investigated the use of carbon ions directly for computed tomography (carbon CT) to create the relative stopping power map of a patient’s body. The Geant4 toolkit was used to perform a Monte Carlo simulation of the carbon ion trajectories, to study their lateral and angular deflections and the most likely paths, using a water phantom. Geant4 was used to create carbonCT projections of a contrast and spatial resolution phantom, with a cone beam of 430 MeV/u carbon ions. The contrast phantom consisted of cranial bone, lung material, and PMMA inserts while the spatial resolution phantom contained bone and lung material inserts with line pair (lp) densities ranging from 1.67 lp cm−1 through 5 lp cm−1. First, the positions of each carbon ion on the rear and front trackers were used for an approximate reconstruction of the phantom. The phantom boundary was extracted from this approximate reconstruction, by using the position as well as angle information from the four tracking detectors, resulting in the entry and exit locations of the individual ions on the phantom surface. Subsequent reconstruction was performed by the iterative algebraic reconstruction technique coupled with total variation minimization (ART-TV) assuming straight line trajectories for the ions inside the phantom. The influence of number of projections was studied with reconstruction from five different sets of projections: 15, 30, 45, 60 and 90. Additionally, the effect of number of ions on the image quality was investigated by reducing the number of ions/projection while keeping the total number of projections at 60. An estimation of carbon ion range using the carbonCT image resulted in improved range prediction compared to the range calculated using a calibration curve.

Characterization of photon-counting multislit breast tomosynthesis.

It has been shown that breast tomosynthesis may improve sensitivity and specificity compared to two-dimensional mammography, resulting in increased detection-rate of cancers or lowered call-back rates. The purpose of this study is to characterize a spectral photon-counting multislit breast tomosynthesis system that is able to do single-scan spectral imaging with multiple collimated x-ray beams. The system differs in many aspects compared to conventional tomosynthesis using energy-integrating flat-panel detectors. The investigated system was a prototype consisting of a dual-threshold photon-counting detector with 21 collimated line detectors scanning across the compressed breast. A review of the system is done in terms of detector, acquisition geometry, and reconstruction methods. Three reconstruction methods were used, simple back-projection, filtered back-projection and an iterative algebraic reconstruction technique. The image quality was evaluated by measuring the modulation transfer-function (MTF), normalized noise-power spectrum, detective quantum-efficiency (DQE), and artifact spread-function (ASF) on reconstructed spectral tomosynthesis images for a total-energy bin (defined by a low-energy threshold calibrated to remove electronic noise) and for a high-energy bin (with a threshold calibrated to split the spectrum in roughly equal parts). Acquisition was performed using a 29 kVp W/Al x-ray spectrum at a 0.24 mGy exposure. The difference in MTF between the two energy bins was negligible, that is, there was no energy dependence on resolution. The MTF dropped to 50% at 1.5 lp/mm to 2.3 lp/mm in the scan direction and 2.4 lp/mm to 3.3 lp/mm in the slit direction, depending on the reconstruction method. The full width at half maximum of the ASF was found to range from 13.8 mm to 18.0 mm for the different reconstruction methods. The zero-frequency DQE of the system was found to be 0.72. The fraction of counts in the high-energy bin was measured to be 59% of the total detected spectrum. Scantimes ranged from 4 s to 16.5 s depending on voltage and current settings. The characterized system generates spectral tomosynthesis images with a dual-energy photon-counting detector. Measurements show a high DQE, enabling high image quality at a low dose, which is beneficial for low-dose applications such as screening. The single-scan spectral images open up for applications such as quantitative material decomposition and contrast-enhanced tomosynthesis.

Lamb-Wave-Based Tomographic Imaging Techniques for Hole-Edge Corrosion Monitoring in Plate Structures.

This study presents a novel monitoring method for hole-edge corrosion damage in plate structures based on Lamb wave tomographic imaging techniques. An experimental procedure with a cross-hole layout using 16 piezoelectric transducers (PZTs) was designed. The A0 mode of the Lamb wave was selected, which is sensitive to thickness-loss damage. The iterative algebraic reconstruction technique (ART) method was used to locate and quantify the corrosion damage at the edge of the hole. Hydrofluoric acid with a concentration of 20% was used to corrode the specimen artificially. To estimate the effectiveness of the proposed method, the real corrosion damage was compared with the predicted corrosion damage based on the tomographic method. The results show that the Lamb-wave-based tomographic method can be used to monitor the hole-edge corrosion damage accurately.

TH‐C‐103‐04: Computed Tomography Image Reconstruction with a Cylindrical Voxelization Method

Purpose: To accelerate the image reconstruction process for iterative algebraic reconstruction techniques using a cylindrical voxelization method. Methods: Iterative image reconstruction has many advantages, such as noise reduction and mitigation of approximate cone reconstruction artifacts, over analytical image reconstruction. But its high computational cost prohibits its use in many clinical applications, such as image guided radiotherapy (IGRT). The system matrix used in iterative reconstruction is often too large to store on the hard drive or in RAM. The cylindrical geometry is a natural fit for the circular scanning trajectory employed in volumetric CT methods such as cone beam computed tomography (CBCT) and tetrahedron beam computed tomography (TBCT). In this study, the iterative simultaneous algebraic reconstruction technique (SART) algorithm using a cylindrical voxelization method is developed for TBCT. The new algorithm takes advantage of the symmetries of the axial scanning geometry used in TBCT and greatly reduces the size of the system matrix. An additional step is needed for interpolating the cylindrical data onto a Cartesian grid. The reconstruction algorithms were tested with numerical phantoms as well as patient CT data. Results: The system matrix with the cylindrical voxelization method is compact and can be easily stored in the computer RAM. SART reconstruction using the cylindrical voxelization method was implemented for the TBCT. The FOM results for the cylindrical voxelization method were slightly better than for the Cartesian voxelization. The cylindrical method was able to provide a speedup of 40 times over the traditional Cartesian method and to provide a reduction of two orders of magnitude in the storage requirement for the system matrix. Conclusion: Image reconstruction using a cylindrical voxelization method was able to provide a significant improvement in the reconstruction speed while maintaining the image quality provided by the conventional reconstruction method using a Cartesian image grid. This work is partially supported by NIH SBIR contract #HHSN261201100045C.

An investigation on attenuation UCT with wave paths enhancement for breast ultrasound

AbstractBreast cancer is one of the most prevalent health hazards in women all over the world. An annual screening test using a diagnostic ultrasound seems to be an appropriate method to handle this physical threat. In this paper, we propose the use of a two‐dimensional reflection‐mode ultrasonic tomography system for a general breast ultrasound investigation enhanced by an echo wave path using a wave reflecting plate. Our virtual imaging setup system is composed of a 128‐element conventional linear array transducer placed on the top of a water bath and a wave reflector attached around this setup. The shape of the wave reflector was designed as a single plate shape, rectangular shape and curve shape. For projection data, we estimate a line integral insonifying data by using a classical integration slope of attenuation coefficient method in which its derivative is independent of transmittance, reflectance, and noise. By imaging a virtual breast phantom model, the reconstruction results obtained from an iterative algebraic reconstruction technique (ART) algorithm show a promising image. The image is not only indicates a tissue characterization from quantitative attenuation data, but also restores an area covered by hard tissue due to an increased wave path direction from our nonrotational imaging system. These results encourage us to implement our concept with a clinical study in the near future. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

A fast multi-resolution approach to tomographic PIV

Tomographic particle image velocimetry (Tomo-PIV) is a recently developed three-component, three-dimensional anemometric non-intrusive measurement technique, based on an optical tomographic reconstruction applied to simultaneously recorded images of the distribution of light intensity scattered by seeding particles immersed into the flow. Nowadays, the reconstruction process is carried out mainly by iterative algebraic reconstruction techniques, well suited to handle the problem of limited number of views, but computationally intensive and memory demanding. The adoption of the multiplicative algebraic reconstruction technique (MART) has become more and more accepted. In the present work, a novel multi-resolution approach is proposed, relying on the adoption of a coarser grid in the first step of the reconstruction to obtain a fast estimation of a reliable and accurate first guess. A performance assessment, carried out on three-dimensional computer-generated distributions of particles, shows a substantial acceleration of the reconstruction process for all the tested seeding densities with respect to the standard method based on 5 MART iterations; a relevant reduction in the memory storage is also achieved. Furthermore, a slight accuracy improvement is noticed. A modified version, improved by a multiplicative line of sight estimation of the first guess on the compressed configuration, is also tested, exhibiting a further remarkable decrease in both memory storage and computational effort, mostly at the lowest tested seeding densities, while retaining the same performances in terms of accuracy.

Development of a GNSS water vapour tomography system using algebraic reconstruction techniques

A GNSS water vapour tomography system developed to reconstruct spatially resolved humidity fields in the troposphere is described. The tomography system was designed to process the slant path delays of about 270 German GNSS stations in near real-time with a temporal resolution of 30 min, a horizontal resolution of 40 km and a vertical resolution of 500 m or better. After a short introduction to the GPS slant delay processing the framework of the GNSS tomography is described in detail. Different implementations of the iterative algebraic reconstruction techniques (ART) used to invert the linear inverse problem are discussed. It was found that the multiplicative techniques (MART) provide the best results with least processing time, i.e., a tomographic reconstruction of about 26,000 slant delays on a 8280 cell grid can be obtained in less than 10 min. Different iterative reconstruction techniques are compared with respect to their convergence behaviour and some numerical parameters. The inversion can be considerably stabilized by using additional non-GNSS observations and implementing various constraints. Different strategies for initialising the tomography and utilizing extra information are discussed. At last an example of a reconstructed field of the wet refractivity is presented and compared to the corresponding distribution of the integrated water vapour, an analysis of a numerical weather model (COSMO-DE) and some radiosonde profiles.

Tomographic particle tracking velocimetry using telecentric imaging

The goal of this article is to discuss 3D Particle Tracking Velocimetry (PTV) in a tomographic reconstructed voxel space with at least doubling the spatial resolution compared to classical 3D PTV. For this purpose, a new tomographic reconstruction technique based on telecentric imaging in combination with the epipolar geometry is presented. The method overcomes the need for memory intensive weighting matrices or cost intensive iterations, which are necessary in iterative algebraic reconstruction techniques. A characteristic of tomographic reconstruction is the reconstruction of ghost particles. As the aim of PTV is the reconstruction of true particle paths, this article focuses on the removal of ghost particles and ghost trajectories. The method is validated via a synthetic turbulent flow field and via the benchmark experiment of a vortex ring.

Is Iterative Reconstruction Ready for MDCT?

Although the very first computed tomographic scanners used the iterative algebraic reconstruction technique, the filtered back-projection (FBP) method soon became the gold standard for computed tomographic reconstruction. Image quality has dramatically improved over the past 30 years thanks to advances in x-ray tubes, detector technologies, and overall system design and integration, as well as refinements to image reconstruction algorithms. Typical effective dose values range from 5 to 7 mSv for chest computed tomography (CT) and from 8 to 12 mSv for abdominal and pelvic CT, compared with 0.02 to 0.2 mSv for chest x-ray, making CT a comparatively high-dose modality.

Emission Spectral Tomography Reconstruction Based on Maximum Entropy Interpolation

To improve the performance of optical computed tomography (OpCT) reconstruction in the case of limited projection views, maximum entropy (ME) algorithms were proposed and can achieve better results than traditional ones. However, in the discrete iterative process of ME, the variables of the iterative function are continuous. Hence, interpolation methods ought to be used to improve the precision of the iterative function values. Here, a sinc function interpolation approach was adopted in ME algorithm (SINCME) and its reconstruction results for OpCT with limited views were studied using four typical phantoms. Compared results with ME without interpolation, traditional medical CT back-projection algorithm (BP), and iterative algorithm algebraic reconstruction technique (ART) showed that the SINCME algorithm achieved the best reconstruction results. In an experiment of emission spectral tomography reconstruction, SINCME was also adopted to calculate the three-dimensional distribution of physical parameters of a candle flame. The studies of both algorithm and experiment demonstrated that SINCME met the demand of limited-view OpCT reconstruction.

A six-dimensional approach to microtexture analysis

A mathematical framework is presented for microtexture derivation and analysis. A six-dimensional space is introduced, defined as the product of direct space (x, y, z) and orientation space, parametrized by the Rodrigues vector (r 1, r 2, r 3). The framework is adapted to diffraction experiments, where orientation information from many grains or deformation microstructures is recorded in parallel, with the prime aim of allowing fast measurements and thus the ability to study microstructure dynamics. Assuming the use of a monochromatic beam the geometry of the projection surfaces is deduced. The derivation of the microtexture is formulated as an inverse problem similar to reconstruction problems in absorption or emission tomography. The applicability of the two major types of reconstruction algorithm, namely transform methods and algebraic methods, is discussed. An explicit solution in terms of the iterative algebraic reconstruction technique algorithm is given. The choice of basis function is discussed. Furthermore, the framework is applied to a number of important simplifying cases. These include grain-by-grain reconstruction, classical pole figure inversion and the generation of three-dimensional maps of non-deformed grains.

Quantitative characterization of microsamples by x‐ray fluorescence tomography

AbstractA novel semi‐quantitative algebraic reconstruction technique for x‐ray fluorescence tomography data sets is described and its use is demonstrated on laser‐ablated resin samples. The experiments were performed at the ID22 beamline of the European Synchrotron Radiation Facility in Grenoble. The projections were collected using a parallel scheme using 22.5 keV x‐rays at 4 µm resolution. The iterative algebraic reconstruction technique was used to calculate an elemental ‘effective’ concentration distribution. This ‘effective’ concentration is introduced for the first time and allows the attenuation of both the incident x‐ray beam and the fluorescence to be included in the reconstruction procedure. Copyright © 2002 John Wiley & Sons, Ltd.

Isotherms in a horizontal differentially heated cavity at intermediate Rayleigh numbers

Rayleigh-Benard convection in a horizontal, differentially heated, high aspect ratio fluid layer is considered. Experiments with three Rayleigh numbers (13900, 34800 and 51800) corresponding to different flow regimes have been performed. The temperature field in the fluid has been recorded in the form of interferometric projections, the fringe patterns representing isotherms. For steady and quasi-steady flow, the three dimensional temperature field has been obtained using an iterative algebraic reconstruction technique. Results show the formation of longitudinal rolls in the cavity at a Rayleigh number of 13900. The flow is unsteady at a Rayleigh number of 34800, but a strong indication of the formation of cubic rolls is seen. At a Rayleigh number of 51800, the flow field is completely unsteady with no discernible pattern. The local ray-averaged Nusselt numbers at each of the heated and cooled walls have been determined at the two lower Rayleigh numbers. The systematic variation of Nusselt number over each of the solid surfaces confirms the flow models proposed at these Rayleigh numbers. The average Nusselt number qualitatively matches the published correlations.

Highlights of Selected Microscopy Research Resource Activities at San Diego

Abstract The intermediate high-voltage electron microscope (IVEM) located at the National Center for Microscopy and Imaging Research at San Diego (NCMIR) can image relatively thick specimens that contain substantial three-dimensional (3-D) structure. Electron tomography is an important tool used at NCMIR for deriving 3-D cellular and subcellular structure from IVEM images. Reconstruction algorithms commonly used in electron tomography include weighted back projection, and iterative algebraic reconstruction techniques such as ART and SIRT. Improvements in reconstruction quality are possible using the iterative algorithms. Because these algorithms are computationally intensive, we have ported them to massively parallel computers at the San Diego Supercomputer Center, reducing the computation time over that required with workstation level machines. The quality of tomographic data for the 3-D reconstruction of biological structures is also being enhanced by NCMIR projects to improve the microscope. We have designed and constructed special electron optics and microscope control systems for the JEOL 4000EX.

A comparison of transform and iterative reconstruction techniques for a volume-imaging PET scanner with a large axial acceptance angle

The authors present the results of comparing 3-D transform and iterative reconstruction methods with measured PET data from the HEAD PENN-PET scanner, which has a very large axial acceptance angle of 27/spl deg/. The algorithms compared are the 3-D-reprojection (3DRP) method, the iterative algebraic reconstruction technique (ART), and the maximum likelihood (ML-EM) algorithm. For the iterative methods, alternatives to the cubic voxel basis functions are also implemented. The comparisons are based on a series of figures of merit, including the point-spread function, the contrast recovery coefficient, and signal-to-noise ratios (SNR). The results show that the ART method implemented with the new basis functions requires only one cycle through the data to produce images with measured SNR values comparable to or better than the 3DRP method. The ML-EM method was run for 10 iterations and produced images with lower SNR values for the test phantom used. With the exception of the ML-EM results, these are in agreement with the authors' previous simulation studies.