Standard Reconstruction Algorithms Research Articles

Evaluation of T2W FLAIR MR image quality using artificial intelligence image reconstruction techniques in the pediatric brain

BackgroundArtificial intelligence (AI) reconstruction techniques have the potential to improve image quality and decrease imaging time. However, these techniques must be assessed for safe and effective use in clinical practice.ObjectiveTo assess image quality and diagnostic confidence of AI reconstruction in the pediatric brain on fluid-attenuated inversion recovery (FLAIR) imaging.Materials and methodsThis prospective, institutional review board (IRB)-approved study enrolled 50 pediatric patients (median age=12 years, Q1=10 years, Q3=14 years) undergoing clinical brain MRI. T2-weighted (T2W) FLAIR images were reconstructed by both standard clinical and AI reconstruction algorithms (strong denoising). Images were independently rated by two neuroradiologists on a dedicated research picture archiving and communication system (PACS) to indicate whether AI increased, decreased, or had no effect on image quality compared to standard reconstruction. Quantitative analysis of signal intensities was also performed to calculate apparent signal to noise (aSNR) and apparent contrast to noise (aCNR) ratios.ResultsAI reconstruction was better than standard in 99% (reader 1, 49/50; reader 2, 50/50) for overall image quality, 99% (reader 1, 49/50; reader 2, 50/50) for subjective SNR, and 98% (reader 1, 49/50; reader 2, 49/50) for diagnostic preference. Quantitative analysis revealed significantly higher gray matter aSNR (30.6±6.5), white matter aSNR (21.4±5.6), and gray-white matter aCNR (7.1±1.6) in AI-reconstructed images compared to standard reconstruction (18±2.7, 14.2±2.8, 4.4±0.8, p<0.001) respectively.ConclusionWe conclude that AI reconstruction improved T2W FLAIR image quality in most patients when compared with standard reconstruction in pediatric patients.

PLIC-Net: A machine learning approach for 3D interface reconstruction in volume of fluid methods

The accurate reconstruction of immiscible fluid–fluid interfaces from the volume fraction field is a critical component of geometric Volume of Fluid methods. A common strategy is the Piecewise Linear Interface Calculation (PLIC), which fits a plane in each mixed-phase computational cell. However, recent work goes beyond PLIC by using two planes or even a paraboloid. To select such planes or paraboloids, complex optimization algorithms as well as carefully crafted heuristics are necessary. Yet, the potential exists for a well-trained machine learning model to efficiently provide broadly applicable solutions to the interface reconstruction problem at lower costs. In this work, the viability of a machine learning approach is demonstrated in the context of a single plane reconstruction. A feed-forward deep neural network is used to predict the normal vector of a PLIC plane given volume fraction and phasic barycenter data in a 3×3×3 stencil. The PLIC plane is then translated in its cell to ensure exact volume conservation. Our proposed neural network PLIC reconstruction (PLIC-Net) is equivariant to reflections about the Cartesian planes. Training data is analytically generated with O(106) randomized paraboloid surfaces, which allows for the sampling a broad range of interface shapes. PLIC-Net is tested in multiphase flow simulations where it is compared to standard LVIRA and ELVIRA reconstruction algorithms, and the impact of training data statistics on PLIC-Net’s performance is also explored. It is found that PLIC-Net greatly limits the formation of spurious planes and generates cleaner numerical break-up of the interface. Additionally, the computational cost of PLIC-Net is lower than that of LVIRA and ELVIRA. These results establish that machine learning is a viable approach to Volume of Fluid interface reconstruction and is superior to current reconstruction algorithms for some cases.

Deep Image Prior-Based PET Reconstruction From Partial Data

In this paper, we propose an unsupervised deep learning method for positron emission tomography reconstruction (PET) from incomplete data. This method utilizes the so-called deep image prior (DIP) as an untrained deep convolutional neural network (CNN) to generate object reconstructions. The main idea is to re-parameterize the image reconstruction problem as a neural network optimization problem. We show that the proposed method effectively addresses the incomplete data reconstruction problem, which otherwise degrades the image resolution and quality of standard reconstruction algorithms. Meanwhile, the proposed method does not require any pre-training procedures, i.e., it is not biased toward any particular dataset. Hence, it has the potential to be used in clinical situations, where training data would be infeasible or prohibitively expensive. The performance of the proposed approach is evaluated with noisy synthetic data based on shepp-logan and brainweb phantoms, and clinical naive rat data. In addition, robustness studies of the approach with respect to regularization parameters are also carried out. We showcase that the proposed method considerably outperforms the state-of-the-art methods, leading to flexible reconstruction from incomplete PET data.

An evaluation of an energy independent CT reconstruction algorithm for use in radiotherapy treatment planning.

Radiotherapy treatment planning relies upon density information provided by CT for accurate dose calculations. Hounsfield units (HUs) are converted to electron/physical density via an energy dependant calibration curve. Multiple curves are required to make full use of the available accelerating potentials (kVp). The curves are bi-linear with a discontinuity occurring at soft-tissue densities. The commercial algorithm, DirectDensityTM (Siemens Healthcare GmbH), constructs a single calibration curve covering all available kVp. This enables the optimisation of the CT image quality, e.g. in terms of contrast, or the reduction of the imaging dose, whilst rendering the radiotherapy treatment dose calculation robust to the energy used to acquire the CT image. We report our investigations on the clinical utilisation of the DirectDensityTM algorithm for radiotherapy treatments, by using all accelerating potentials, i.e. from 70 kVp up to 140 kVp, available at our CT treatment simulator, in contrast to previous studies that were limited to accelerating potentials spanning a subset of the available kVp. The DirectDensityTM (DD) reconstruction algorithm available on a SOMATOM go.Open Pro CT scanner (Siemens Healthineers) was evaluated using the RayStation v. 9 treatment planning system (RaySearch Laboratories, Stockholm, Sweden) and a CIRS Model 002LFC IMRT Thorax Phantom (SunNuclear, Melbourne, FL), which was imaged at all available kVp with clinical protocols corresponding to various anatomical sites. The DD images were compared to those with the standard reconstruction algorithm acquired only at 120 kVp, as per our routine clinical practice. The effect of increasing kVp on HU is investigated for relevant tissue substitutes. In addition, a dosimetric comparison is performed for a VMAT plan technique with 6 MV X-rays using retrospective patient CT data sets representing four anatomical sites (pelvis, thorax, brain and "head and neck") with five patients for each site. The original dose distributions were calculated on images acquired at 120 kVp using the standard clinical iterative reconstruction (Qr40) and compared with dose distributions recalculated on images reconstructed with the new DD (Sm40) algorithm. The maximum difference for radiotherapy doses calculated using images of the phantom reconstructed with Qr40 (120 kVp) or DD (all available kVp) was 0.73%. The patient plans on the anatomically representative sites studied here showed a mean PTV dose difference of -0.2% (s.d. 0.7) for D99%, -0.4% (s.d. 0.4%) for D50% and -0.3% (s.d. 0.4%) for D2%. Incidentally, we found a previously unreported decrease in HU, mostly notable for bone type inserts (~34 HU (cortical bone)), at 110 kVp for the DD reconstructed images. The effect was not noted for the standard Qr40 reconstructions. DD has a minimal dosimetric impact in the dose calculations for radiotherapy treatments and could be implemented with existing clinical workflows. Attention should be paid to the HU values for images acquired at 110 kVp (DD algorithm), which warrants further investigation. This is the first paper where DD was evaluated at all available kVp, leading to the incidental discovery of abnormal HU values at 110 kVp for this algorithm.

Deep learning reconstruction vs standard reconstruction for abdominal CT: the influence of BMI.

This study aimed to evaluate the image quality and lesion conspicuity of the deep learning image reconstruction (DLIR) algorithm compared with standard image reconstruction algorithms on abdominal enhanced computed tomography (CT) scanning with a wide range of body mass indexes (BMIs). A total of 112 participants who underwent contrast-enhanced abdominal CT scans were divided into three groups according to BMIs: the 80-kVp group (BMI ≤ 23.9kg/m2), 100-kVp group (BMI 24-28.9kg/m2), and 120-kVp group (BMI ≥ 29kg/m2). All images were reconstructed using filtered back projection (FBP), adaptive statistical iterative reconstruction-V of 50% level (IR), and DLIR at low, medium, and high levels (DL, DM, and DH, respectively). Subjective noise, artifact, overall image quality, and low- and high-contrast hepatic lesion conspicuity were all graded on a 5-point scale. The CT attenuation value (in HU), image noise, and contrast-to-noise ratio (CNR) were quantified and compared. DM and DH improved the qualitative and quantitative parameters compared with FBP and IR for all three BMI groups. DH had the lowest image noise and highest CNR value, while DM had the highest subjective overall image quality and low- and high-contrast lesion conspicuity scores for the three BMI groups. Based on the FBP, the improvement in image quality and lesion conspicuity of DM and DH images was greater in the 80-kVp group than in the 100-kVp and 120-kVp groups. For all BMIs, DLIR improves both image quality and hepatic lesion conspicuity, of which DM would be the best choice to balance both. The study suggests that utilizing DLIR, particularly at the medium level, can significantly enhance image quality and lesion visibility on abdominal CT scans across a wide range of BMIs. • DLIR improved the image quality and lesion conspicuity across a wide range of BMIs. • DLIR at medium level had the highest subjective parameters and lesion conspicuity scores among all reconstruction levels. • On the basis of the FBP, the 80-kVp group had improved image quality and lesion conspicuity more than the 100-kVp and 120-kVp groups.

Validation of semi-analytical, semi-empirical covariance matrices for two-point correlation function for early DESI data

ABSTRACT We present an extended validation of semi-analytical, semi-empirical covariance matrices for the two-point correlation function (2PCF) on simulated catalogs representative of luminous red galaxies (LRGs) data collected during the initial 2 months of operations of the Stage-IV ground-based Dark Energy Spectroscopic Instrument (DESI). We run the pipeline on multiple effective Zel’dovich (EZ) mock galaxy catalogs with the corresponding cuts applied and compare the results with the mock sample covariance to assess the accuracy and its fluctuations. We propose an extension of the previously developed formalism for catalogs processed with standard reconstruction algorithms. We consider methods for comparing covariance matrices in detail, highlighting their interpretation and statistical properties caused by sample variance, in particular, non-trivial expectation values of certain metrics even when the external covariance estimate is perfect. With improved mocks and validation techniques, we confirm a good agreement between our predictions and sample covariance. This allows one to generate covariance matrices for comparable data sets without the need to create numerous mock galaxy catalogs with matching clustering, only requiring 2PCF measurements from the data itself. The code used in this paper is publicly available at https://github.com/oliverphilcox/RascalC.

Resolution-enhanced reflection ptychography with axial distance calibration

Reflection ptychography is a lens-free computational imaging technology that operates by a coherent beam illuminated to a specimen inclined to the optical axis, a set of distorted diffracted patterns is captured on the far-filed plane. This tilted distortion results from the tilt angle of illumination in a reflection mode as a non-coplanar scattering geometry, which mismatches the Fourier transformed distribution and blocks the operation of FFT (fast Fourier Transform) in standard ptychographic reconstruction algorithms. This distortion can be corrected by performing a coordinate transform from the tilted geometry combined with knowledge of the axial distance of diffraction and the radiation wavelength. However, practically the axial distance usually cannot be measured with accuracy instead of a rough estimate, which may disrupt the correction from distortion, and degrade the reconstructed quality significantly. Here, we propose and analyze for the first time a straightforward axial distance calibration approach to improve the reconstructed resolution in reflection ptychography. Firstly, a uniform Fraunhofer-shaped diffraction is transformed from the tilted distorted scatterings by a coordinate transform at a coarse estimate of axial distance. Then, the axial distance error can be precisely corrected from the periodic unwrapped phase of the retrieved probe by ptychography. Finally, a resolution-enhanced reconstruction is generated from the reflection ptychography with axial distance calibration. A group of proof-of-concept experiments was performed for verification. Results demonstrate that the axial distance can be precisely calibrated within a fraction of a millimeter. Moreover, the reconstructed resolution is remarkably improved (2 times improvement in our experimental tests) by utilizing the proposed approach in reflection ptychography.

The application value of deep learning image reconstruction on improving image quality and evaluating the Qanadli embolism index of dual low-dose CT pulmonary angiography

Objective: To compare the image quality and Qanadli embolism index between deep learning image reconstruction (DLR) and adaptive statistical iterative reconstruction-veo (ASiR-V) in dual low-dose CT pulmonary angiography (CTPA) with low contrast agent dose and low radiation dose. Methods: Eighty-eight patients who underwent dual low-dose CTPA in the radiology department of the affiliated hospital of Xuzhou Medical University from October 2020 to March 2021 were retrospectively analyzed, including 44 males and 44 females, aged from 11 to 87 years (61±15 years). The CTPA examination were performed using 80 kV tube voltage and 20 ml contrast agent. The raw data were reconstructed using standard kernel DLR high level (DL-H) and ASiR-V reconstruction, respectively. The patients were divided into standard kernel DL-H group (n=88, 33 cases of positive embolism) and ASiR-V group (n=88, 36 cases of positive embolism). The CT value, image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), subjective image quality score, Qanadli embolism index, positive rate and positive Qanadli embolism index were compared between the two groups. Results: There were no statistically significant differences in CT values of the main pulmonary artery, the right pulmonary artery and the left pulmonary artery between the standard kernel DL-H group and ASiR-V group [(405.8±111.7) vs (404.0±112.0) HU, (412.9±113.1) vs (411.5±112.2) HU, (418.1±119.9) vs (415.4±118.0) HU, respectively;all P>0.05)]. The image noise of the main pulmonary artery, the right pulmonary artery and the left pulmonary artery in the standard kernel DL-H group was significantly lower than the ASiR-V group(16.6±4.7 vs 28.1±4.8, 18.3±6.1 vs 29.8±4.9, 17.6±5.6 vs 28.4±4.7, respectively;all P<0.001). The SNR and CNR of the main pulmonary artery, the right pulmonary artery and the left pulmonary artery in the standard kernel DL-H group were significantly higher than the ASiR-V group(SNR: 25.5±7.1 vs 14.5±3.9, 23.9±7.2 vs 13.9±3.4, 24.9±7.4 vs 14.8±4.1, CNR: 21.6±6.6 vs 12.3±3.9, 20.2±6.7 vs 11.8±3.4, 21.2±6.9 vs 12.6±4.1, respectively;all P<0.001). The subjective image quality score of the standard kernel DL-H group was significantly higher than the ASiR-V group (4.6 vs 3.8, P<0.001). There were no significant difference in the Qanadli embolism index, positive rate and positive Qanadli embolism index between the two groups (all P>0.05). Conclusion: Compared with ASiR-V reconstruction algorithms group, standard kernel DL-H reconstruction algorithms can significantly improve the image quality of dual low-dose CTPA.

Potentials of computer simulation of lung tumors in comparison with &lt;sup&gt;99m&lt;/sup&gt;Тс-MIBI SPECT/CT data

The aim of the study was to develop and validate a software package (SP) for computer simulation of the procedure for examining patients with lung cancer by SPECT/CT and assessing the accuracy of reconstruction of tumor lesions.Material and Methods. Lung scintigraphy for a patient with peripheral squamous cell carcinoma of the upper lobe of the right lung was performed using a two-detector gamma camera GE Discovery NM/CT 670 DR (USA) with high-resolution collimators for an energy of 140 KeV and a radiopharmaceutical (RP) 99mTc-Technetril (MIBI, Diamed, Moscow). The data obtained were subjected to computer processing using a specialized Xeleris 4.0 system from GE (USA). The SP included a program for generating a voxel phantom (“virtual patient”), a program for modeling the “raw” data acquisition (“virtual tomograph”) and an image reconstruction based on the OSEM algorithm (Ordered Subset Expectation Maximization). In order to validate the created SP, computer simulation of the above clinical case was performed. The semi-quantitative comparative image analysis was based on a tumor/background score.Results. There was a good correlation between clinical “raw” data recorded from a real patient and projection data calculated by the Monte Carlo method from a “virtual patient”. The results of the comparative analysis showed that the tumor/background assessment was underestimated in the reconstructed images.Conclusion. The problem of the accuracy of the tumor lesions reconstruction by using standard OSEM reconstruction algorithms has not been studied. This issue is important in the management of patients with tumor lesions of the lungs and requires study and systematization. The SP will be used in further studies to analyze errors and artifacts in images of tumor lesions, as well as to develop approaches to overcome them.

Development of a vertex finding algorithm using Recurrent Neural Network

Deep learning is a rapidly-evolving technology with possibility to significantly improve physics reach of collider experiments. In this study we developed a novel algorithm of vertex finding for future lepton colliders such as the International Linear Collider. We deploy two networks; one is simple fully-connected layers to look for vertex seeds from track pairs, and the other is a customized Recurrent Neural Network with an attention mechanism and an encoder-decoder structure to associate tracks to the vertex seeds. The performance of the vertex finder is compared with the standard ILC reconstruction algorithm.

Deep learning for terahertz image denoising in nondestructive historical document analysis

Historical documents contain essential information about the past, including places, people, or events. Many of these valuable cultural artifacts cannot be further examined due to aging or external influences, as they are too fragile to be opened or turned over, so their rich contents remain hidden. Terahertz (THz) imaging is a nondestructive 3D imaging technique that can be used to reveal the hidden contents without damaging the documents. As noise or imaging artifacts are predominantly present in reconstructed images processed by standard THz reconstruction algorithms, this work intends to improve THz image quality with deep learning. To overcome the data scarcity problem in training a supervised deep learning model, an unsupervised deep learning network (CycleGAN) is first applied to generate paired noisy THz images from clean images (clean images are generated by a handwriting generator). With such synthetic noisy-to-clean paired images, a supervised deep learning model using Pix2pixGAN is trained, which is effective to enhance real noisy THz images. After Pix2pixGAN denoising, 99% characters written on one-side of the Xuan paper can be clearly recognized, while 61% characters written on one-side of the standard paper are sufficiently recognized. The average perceptual indices of Pix2pixGAN processed images are 16.83, which is very close to the average perceptual index 16.19 of clean handwriting images. Our work has important value for THz-imaging-based nondestructive historical document analysis.

Iterative Reconstruction: State-of-the-Art and Future Perspectives.

Image reconstruction processing in computed tomography (CT) has evolved tremendously since its creation, succeeding at optimizing radiation dose while maintaining adequate image quality. Computed tomography vendors have developed and implemented various technical advances, such as automatic noise reduction filters, automatic exposure control, and refined imaging reconstruction algorithms.Focusing on imaging reconstruction, filtered back-projection has represented the standard reconstruction algorithm for over 3 decades, obtaining adequate image quality at standard radiation dose exposures. To overcome filtered back-projection reconstruction flaws in low-dose CT data sets, advanced iterative reconstruction algorithms consisting of either backward projection or both backward and forward projections have been developed, with the goal to enable low-dose CT acquisitions with high image quality. Iterative reconstruction techniques play a key role in routine workflow implementation (eg, screening protocols, vascular and pediatric applications), in quantitative CT imaging applications, and in dose exposure limitation in oncologic patients.Therefore, this review aims to provide an overview of the technical principles and the main clinical application of iterative reconstruction algorithms, focusing on the strengths and weaknesses, in addition to integrating future perspectives in the new era of artificial intelligence.

Inter-observer agreement and image quality of model-based algorithm applied to the Coronary Artery Disease-Reporting and Data System score

PurposeTo evaluate the inter-observer agreement of the CAD-RADS reporting system and compare image quality between model-based iterative reconstruction algorithm (MBIR) and standard iterative reconstruction algorithm (IR) of low-dose cardiac computed tomography angiography (CCTA).MethodsOne-hundred-sixty patients undergone a 256-slice MDCT scanner using low-dose CCTA combined with prospective ECG-gated techniques were enrolled. CCTA protocols were reconstructed with both MBIR and IR. Each study was evaluated by two readers using the CAD-RADS lexicon. Vessels enhancement, image noise, signal-to-noise (SNR), and contrast-to-noise (CNR) were computed in the axial native images, and inter-observer agreement was assessed. Radiation dose exposure as dose–length product (DLP) and effective dose were finally reported.ResultsThe reliability analysis between the two readers was almost perfect for all CAD-RADS standard categories. Moreover, a significantly higher value of subjective qualitative analysis, SNR, and CNR in MBIR images compared to IR were found, due to a lower noise level (all p < 0.05). The mean DLP measured was 63.9 mGy*cm, and the mean effective dose was 0.9 mSv.ConclusionInter-observer agreement of CAD-RADS was excellent confirming the importance, the feasibility, and the reproducibility of the CAD-RADS scoring system for CCTA. Moreover, lower noise and higher image quality with MBIR compared to IR were found.Implications for practiceMBIR, by reducing noise and improving image quality, can help a better assessment of CAD-RADS, in comparison with standard IR algorithm.

Atom Probe Analysis of Nanoparticles Through Pick and Coat Sample Preparation.

The ability to analyze nanoparticles in the atom probe has often been limited by the complexity of the sample preparation. In this work, we present a method to lift–out single nanoparticles in the scanning electron microscope. First, nanoparticles are dispersed on a lacey carbon grid, then positioned on a sharp substrate tip and coated on all sides with a metallic matrix by physical vapor deposition. Compositional and structural insights are provided for spherical gold nanoparticles and a segregation of silver and copper in silver copper oxide nanorods is shown in 3D atom maps. Using the standard atom probe reconstruction algorithm, data quality is limited by typical standard reconstruction artifacts for heterogeneous specimens (trajectory aberrations) and the choice of suitable coatings for the particles. This approach can be applied to various unsupported free-standing nanoparticles, enables preselection of particles via correlative techniques, and reliably produces well-defined structured samples. The only prerequisite is that the nanoparticles must be large enough to be manipulated, which was done for sizes down to ~50 nm.

Wall Shear Stress Measurements Using A MTV-Based Plenoptic Microscope

In this contribution, we describe the first ever microscopic Molecular Tagging Velocimetry (MTV) measurements using Integral Imaging. This technique permits the measurement of the 3D-2C velocity field in an axi-symmetric stagnation jet and observed profiles agree well with analytical predictions. Standard reconstruction algorithms are compared against more advanced deconvolution algorithms and impressive improvements in reconstructed probe resolution are obtained. Integral imaging is compared with a scanning confocal microscope which serves as a basis of comparison and good agreement is achieved. The flow facility was designed for a low Reynolds number, but with an adjustable viscous length scale comparable with higher Reynolds number flows. Our instrument proves capable of resolving viscous length scales on the order of 30 micrometers which can be improved by increasing the numerical aperture of the microscope objective.

Accuracy and Reproducibility of Myocardial Blood Flow Quantification by Single Photon Emission Computed Tomography Imaging in Patients With Known or Suspected Coronary Artery Disease.

Single photon emission computed tomography (SPECT) has limited ability to identify multivessel and microvascular coronary artery disease. Gamma cameras with cadmium zinc telluride detectors allow the quantification of absolute myocardial blood flow (MBF) and myocardial flow reserve (MFR). However, evidence of its accuracy is limited, and of its reproducibility is lacking. We aimed to validate 99mTc-sestamibi SPECT MBF and MFR using standard and spline-fitted reconstruction algorithms compared with 13N-ammonia positron emission tomography in a cohort of patients with known or suspected coronary artery disease and to evaluate the reproducibility of this technique. Accuracy was assessed in 34 participants who underwent dynamic 99mTc-sestamibi SPECT and 13N-ammonia positron emission tomography and reproducibility in 14 participants who underwent 2 99mTc-sestamibi SPECT studies, all within 2 weeks. A rest/pharmacological stress single-day SPECT protocol was performed. SPECT images were reconstructed using a standard ordered subset expectation maximization (OSEM) algorithm with (N=21) and without (N=30) application of spline fitting. SPECT MBF was quantified using a net retention kinetic model' and MFR was derived as the stress/rest MBF ratio. SPECT global MBF with splines showed good correlation with 13N-ammonia positron emission tomography (r=0.81, P<0.001) and MFR estimates (r=0.74, P<0.001). Correlations were substantially weaker for standard reconstruction without splines (r=0.61, P<0.001 and r=0.34, P=0.07, for MBF and MFR, respectively). Reproducibility of global MBF estimates with splines in paired SPECT scans was good (r=0.77, P<0.001), while ordered subset expectation maximization without splines led to decreased MBF (r=0.68, P<0.001) and MFR correlations (r=0.33, P=0.3). There were no significant differences in MBF or MFR between the 2 reproducibility scans independently of the reconstruction algorithm (P>0.05 for all). MBF and MFR quantification using 99mTc-sestamibi cadmium zinc telluride SPECT with spatiotemporal spline fitting improved the correlation with 13N-ammonia positron emission tomography flow estimates and test/retest reproducibility. The use of splines may represent an important step toward the standardization of SPECT flow estimation.

Neural network regularization in the problem of few-view computed tomography

The computed tomography allows to reconstruct the inner morphological structure of an object without physical destructing. The accuracy of digital image reconstruction directly depends on the measurement conditions of tomographic projections, in particular, on the number of recorded projections. In medicine, to reduce the dose of the patient load there try to reduce the number of measured projections. However, in a few-view computed tomography, when we have a small number of projections, using standard reconstruction algorithms leads to the reconstructed images degradation. The main feature of our approach for few-view tomography is that algebraic reconstruction is being finalized by a neural network with keeping measured projection data because the additive result is in zero space of the forward projection operator. The final reconstruction presents the sum of the additive calculated with the neural network and the algebraic reconstruction. First is an element of zero space of the forward projection operator. The second is an element of orthogonal addition to the zero space. Last is the result of applying the algebraic reconstruction method to a few-angle sinogram. The dependency model between elements of zero space of forward projection operator and algebraic reconstruction is built with neural networks. It demonstrated that realization of the suggested approach allows achieving better reconstruction accuracy and better computation time than state-of-the-art approaches on test data from the Low Dose CT Challenge dataset without increasing reprojection error.

Superresolution Reconstruction of Remote Sensing Image Based on Generative Adversarial Network

To recreate high-resolution, more detailed remote sensing images from existing low-resolution photos, this technique is known as remote sensing image superresolution reconstruction, and it has numerous uses. As an important research hotspot of neural networks, generative adversarial network (GAN) has made outstanding progress for image superresolution reconstruction. It solves the computational complexity and low reconstructed image quality of standard superresolution reconstruction algorithms. This research offers a superresolution reconstruction strategy with a self-attention generative adversarial network to improve the quality of reconstructed superresolution remote sensing images. The self-attention strategy as well as residual module is utilized to build a generator in this model that transforms low-resolution remote sensing images into superresolution ones. It aims to determine the discrepancy between a reconstructed picture and a true picture by using a deep convolutional network as a discriminator. For the purpose of enhancing the accuracy, content loss is used. This is done to obtain accurate detail reconstruction. According to the findings of the experiments, this approach is capable of regenerating higher-quality images.

Static coded illumination strategies for low-dose x-ray material decomposition.

Static coded aperture x-ray tomography was introduced recently where a static illumination pattern is used to interrogate an object with a low radiation dose, from which an accurate 3D reconstruction of the object can be attained computationally. Rather than continuously switching the pattern of illumination with each view angle, as traditionally done, static code computed tomography (CT) places a single pattern for all views. The advantages are many, including the feasibility of practical implementation. This paper generalizes this powerful framework to develop single-scan dual-energy coded aperture spectral tomography that enables material characterization at a significantly reduced exposure level. Two sensing strategies are explored: rapid kV switching with a single-static block/unblock coded aperture, and coded apertures with non-uniform thickness. Both systems rely on coded illumination with a plurality of x-ray spectra created by kV switching or 3D coded apertures. The structured x-ray illumination is projected through the objects of interest and measured with standard x-ray energy integrating detectors. Then, based on the tensor representation of projection data, we develop an algorithm to estimate a full set of synthesized measurements that can be used with standard reconstruction algorithms to accurately recover the object in each energy channel. Simulation and experimental results demonstrate the effectiveness of the proposed cost-effective solution to attain material characterization in low-dose dual-energy CT.

Human in-vivo magnetic resonance current density imaging of the brain by optimizing head tissue conductivities

Abstract Introduction: Magnetic resonance current density imaging (MRCDI) aims to reconstruct the current flow of transcranial electrical stimulation (TES) in the brain from MR-measurements of the current-induced magnetic field Bz. Aim: We test the performance of a standard reconstruction algorithm (“projected current density algorithm”, PCD, Jeong et al. 2014) for human brain data. We compare it with current flow simulations using personalized head models. Methods:1. We generated ground-truth data for the TES current flow and Bz-field using a detailed head model and SimNIBS (www.simnibs.org). We applied the PCD algorithm to the Bz -field and quantified the reconstruction performance by comparison with the ground-truth current flow. We additionally compared the PCD results with simulations using a simple head model (“3c” with scalp, bone and a homogeneous intracranial compartment). 2. We reconstructed the current flow from in-vivo MRCDI data (Göksu et al, 2018) with the PCD algorithm. We also used head models of different complexities (“3c” and “4c”: scalp, skull, CSF & brain) and optimized their conductivities to minimize the root-mean-square difference between the measured and simulated Bz. Results: 1. For simulated Bz data, the PCD algorithm only coarsely reconstructed the true current flow. Even the simple head model performed better. 2. For measured Bz data, current flows obtained with personalized head models and fitted conductivities explained the measurements better than the current flow reconstructed with the PCD algorithm. This was already the case for the simple head model (3c). The more detailed model (4c) resulted in further statistically significant improvements. However, for all models, the unexplained variance stayed above the noise floor, indicating remaining differences to unknown true current flow. Conclusions: The PCD algorithm has low accuracy for MRCDI data of the brain. However, MRCDI is useful for evaluations and improvements of current flow simulations with anatomically detailed personalized head models. Keywords: magnetic resonance current density imaging, personalized electric field calculation, dose control, transcranial electrical stimulation