Simulated X-ray Images Research Articles

Simulation of X-ray projections on GPU: Benchmarking gVirtualXray with clinically realistic phantoms

Background and objectives: This study provides a quantitative comparison of images created using gVirtualXray (gVXR) to both Monte Carlo (MC) and real images of clinically realistic phantoms. gVirtualXray is an open-source framework that relies on the Beer-Lambert law to simulate X-ray images in realtime on a graphics processor unit (GPU) using triangular meshes.Methods: Images are generated with gVirtualXray and compared with a corresponding ground truth image of an anthropomorphic phantom: (i) an X-ray projection generated using a Monte Carlo simulation code, (ii) real digitally reconstructed radiographs (DRRs), (iii) computed tomography (CT) slices, and (iv) a real radiograph acquired with a clinical X-ray imaging system. When real images are involved, the simulations are used in an image registration framework so that the two images are aligned.Results: The mean absolute percentage error (MAPE) between the images simulated with gVirtualXray and MC is 3.12%, the zero-mean normalised cross-correlation (ZNCC) is 99.96% and the structural similarity index (SSIM) is 0.99. The run-time is 10 days for MC and 23 ms with gVirtualXray. Images simulated using surface models segmented from a CT scan of the Lungman chest phantom were similar to (i) DRRs computed from the CT volume and (ii) an actual digital radiograph. CT slices reconstructed from images simulated with gVirtualXray were comparable to the corresponding slices of the original CT volume.Conclusions: When scattering can be ignored, accurate images that would take days using MC can be generated in milliseconds with gVirtualXray. This speed of execution enables the use of repetitive simulations with varying parameters, e.g. to generate training data for a deep-learning algorithm, and to minimise the objective function of an optimisation problem in image registration. The use of surface models enables the combination of X-ray simulation with real-time soft-tissue deformation and character animation, which can be deployed in virtual reality applications.

Image Separation With Side Information: A Connected Auto-Encoders Based Approach

X-radiography (X-ray imaging) is a widely used imaging technique in art investigation. It can provide information about the condition of a painting as well as insights into an artist's techniques and working methods, often revealing hidden information invisible to the naked eye. X-radiograpy of double-sided paintings results in a mixed X-ray image and this paper deals with the problem of separating this mixed image. Using the visible color images (RGB images) from each side of the painting, we propose a new Neural Network architecture, based upon 'connected' auto-encoders, designed to separate the mixed X-ray image into two simulated X-ray images corresponding to each side. This connected auto-encoders architecture is such that the encoders are based on convolutional learned iterative shrinkage thresholding algorithms (CLISTA) designed using algorithm unrolling techniques, whereas the decoders consist of simple linear convolutional layers; the encoders extract sparse codes from the visible image of the front and rear paintings and mixed X-ray image, whereas the decoders reproduce both the original RGB images and the mixed X-ray image. The learning algorithm operates in a totally self-supervised fashion without requiring a sample set that contains both the mixed X-ray images and the separated ones. The methodology was tested on images from the double-sided wing panels of the Ghent Altarpiece, painted in 1432 by the brothers Hubert and Jan van Eyck. These tests show that the proposed approach outperforms other state-of-the-art X-ray image separation methods for art investigation applications.

Calibration of 3D scan trajectories for an industrial computed tomography setup with 6-DOF object manipulator system using a single sphere

In industrial x-ray computed tomography (CT), the application of more complex scan paths in comparison to the typical circular trajectory ( rotation of the measurement object) can extend the potential of CT. One way to enable such 3D scan trajectories is to use a 6-degrees-of-freedom (DOF) object manipulator system. In our case, a hexapod is mounted on top of the rotary table of a commercial CT scanner. This allows for adaptive tilting of the measurement object during the scan. For high accuracy, the geometry calibration of such setups is typically done using the x-ray projections of a calibrated multi-sphere object. Contrary to this, here, we demonstrate a procedure that is based on only a single sphere and can therefore experimentally be implemented with low effort. Using the intrinsic geometry parameters of the CT device as prior information, the hexapod coordinate system with respect to the CT machine coordinate system is determined by means of a one-step optimization approach. The resulting parameters are used to calculate projection matrices that enable the volume reconstruction for 3D scan trajectories. The method is validated using simulated x-ray images and experimental investigations including dimensional measurements. For the used setup, geometric measurement results for 3D scan trajectories that are calibrated with the presented method show in sum increased errors compared to the circular scans. A limited pose accuracy of the manipulator system is discussed as a potential cause. The results nevertheless indicate that the presented method is generally feasible for dimensional CT measurements provided that the pose accuracy is sufficient. The calibration procedure can therefore be a low-cost and easier to implement alternative compared to trajectory calibration methods based on multi-sphere objects, but with a tendency towards lower measurement accuracy. The methodology can in principle be transferred to different setups with 6-DOF manipulator systems, e.g. C-arm CT devices with a robot arm.

Computational models of direct and indirect X-ray breast imaging detectors for in silico trials.

To facilitate in silico studies that investigate digital mammography (DM) and breast tomosynthesis (DBT), models replicating the variety in imaging performance of the DM and DBT systems, observed across manufacturers areneeded. The main purpose of this work is to develop generic physics models for direct and indirect detector technology used in commercially available systems, with the goal of making them available open source to manufacturers to further tweak and develop the exact in silico replicas of theirsystems. We recently reported on an in silico version of the SIEMENS Mammomat Inspiration DM/DBT system using an open-source GPU-accelerated Monte Carlo x-ray imaging simulation code (MC-GPU). We build on the previous version of the MC-GPU codes to mimic the imaging performances of two other Food and Drug Administration (FDA)-approved DM/DBT systems, such as Hologic Selenia Dimensions (HSD) and the General Electric Senographe Pristina (GSP) systems. In this work, we developed a hybrid technique to model the optical spread and signal crosstalk observed in the GSP and HSD systems. MC simulations are used to track each x-ray photon till its first interaction within the x-ray detector. On the other hand, the signal spread in the x-ray detectors is modeled using previously developed analytical equations. This approach allows us to preserve the modeling accuracy offered by MC methods in the patient body, while speeding up secondary carrier transport (either electron-hole pairs or optical photons) using analytical equationsin the detector. The analytical optical spread model for the indirect detector includes the depth-dependent spread and collection of optical photons and relies on a pre-computed set of point response functions that describe the optical spread as a function of depth. To understand the capabilities of the computational x-ray detector models, we compared image quality metrics like modulation transfer function (MTF), normalized noise power spectrum (NNPS), and detective quantum efficiency (DQE), simulated with our models against measured data. Please note that the purpose of these comparisons with measured data would be to gauge if the model developed as part of this work could replicate commercially used direct and indirect technology in general and not to achieve perfect fits with measureddata. We found that the simulated image quality metrics such as MTF, NNPS, and DQE were in reasonable agreement with experimental data. To demonstrate the imaging performance of the three DM/DBT systems, we integrated the detector models with the VICTRE pipeline and simulated DM images of a fatty breast model containing a spiculated mass and a calcium oxalate cluster. In general, we found that the images generated using the indirect model appeared more blurred with a different noise texture and contrast as compared to the systems with directdetectors. We have presented computational models of three commercially available FDA-approved DM/DBT systems, which implement both direct and indirect detector technology. The updated versions of the MC-GPU codes that can be used to replicate three systems are available in open source format throughGitHub.

Comparison of clinical effect of MR imaging combined with simulated X-ray and magnetic resonance imaging in patients with benign and malignant bone tumors

ObjectiveTo explore the differential diagnosis results of MRI combined with simulated X-ray in patients with benign or malignant bone tumors. MethodsA total of 143 patients with typical benign or malignant skeletal system tumors were divided into benign (n = 78) and malignant groups (n = 65). They underwent simulated X-ray and MRI examinations before surgery. The imaging results were retrospectively analyzed, and the occurrence rates of various MRI signs and the detection accuracy of simulated X-ray examination in the two groups were calculated. Pathological diagnosis was used as the gold standard. The diagnosis of MRI examination and the combined MRI and X-ray examination for malignant tumor was compared and analyzed. Slope (enhancement slope value) and DER (centripetal enhancement rate) were calculated. The diagnostic effects of each critical point slope on skeletal system malignant tumors were analyzed. ResultsMRI signs showed that there were 107 cases of abnormal bone signals accompanied by solid soft tissue mass, 97.20% in the malignant group and 2.80% in the benign group (P < 0.05). There were 164 cases of peripheral tumor edema, with 68.29% in the malignant group and 31.71% in the benign group. There were 76 cases of radial periosteum reaction and periosteum triangle, with 94.74% in the malignant group and 5.26% in the benign group (P < 0.05). There were 11 cases of bone tumor in the malignant group. The accuracy of MRI combined with simulated X-ray was significantly higher than that of the two methods alone. There was a significant difference between malignant group and benign group in MRI detection slope value (P < 0.05). The slope value of different critical points had different effects in differentiating the characteristics of skeletal system tumors (P < 0.05). ConclusionMRI combined with simulated X-ray examination has a high accuracy rate in differentiating benign and malignant bone tumors. MRI signs showed abnormal signals in bone, accompanied by solid soft tissue mass, radial periosteum reaction and periosteum triangle, which can highly indicate malignant bone tumors.

Generative Adversarial Network (GAN) for Automatic Reconstruction of the 3D Spine Structure by Using Simulated Bi-Planar X-ray Images.

In this study, we modified the previously proposed X2CT-GAN to build a 2Dto3D-GAN of the spine. This study also incorporated the radiologist’s perspective in the adjustment of input signals to prove the feasibility of the automatic production of three-dimensional (3D) structures of the spine from simulated bi-planar two-dimensional (2D) X-ray images. Data from 1012 computed tomography (CT) studies of 984 patients were retrospectively collected. We tested this model under different dataset sizes (333, 666, and 1012) with different bone signal conditions to observe the training performance. A 10-fold cross-validation and five metrics—Dice similarity coefficient (DSC) value, Jaccard similarity coefficient (JSC), overlap volume (OV), and structural similarity index (SSIM)—were applied for model evaluation. The optimal mean values for DSC, JSC, OV, SSIM_anteroposterior (AP), and SSIM_Lateral (Lat) were 0.8192, 0.6984, 0.8624, 0.9261, and 0.9242, respectively. There was a significant improvement in the training performance under empirically enhanced bone signal conditions and with increasing training dataset sizes. These results demonstrate the potential of the clinical implantation of GAN for automatic production of 3D spine images from 2D images. This prototype model can serve as a foundation in future studies applying transfer learning for the development of advanced medical diagnostic techniques.

CAD-based x-ray CT calibration and error compensation

Artifacts due to imperfect determination of the scanner geometry, beam hardening (BH), and diffuse Compton scattering, limit the quantitative exploitation of radiographs or tomographies for non-destructive evaluation. Exploiting the computer-aided design (CAD) model of an industrial part, a methodology is proposed to refine the estimation of the computed tomography (CT)-scanner geometry up to a scale factor, to correct or account for artifacts, and to assess the metrology of the part. A projective model describing the formation of x-ray images in CT-scanners is first introduced. The optimal parameters of the projective model are identified using a novel CAD-based calibration method that relies on the registration of simulated projections onto experimental ones. A metrological analysis based on the comparison between acquired and simulated x-ray images is proposed. A turbine blade, for which an automatic inspection procedure from few views is under development, is used as an example to illustrate the proposed methodology. The parametrization accounts for the refinement of the projection geometry, the calibration of BH, and the estimation of scattering. It is shown that, using the proposed procedure, the differences between acquired and simulated radiographic images are significantly reduced, indicating that the optimal parameters are properly identified. These differences are then exploited to detect flaws in the part.

Autoregression and Structured Low-Rank Modeling of Sinogram Neighborhoods.

Sinograms are commonly used to represent the raw data from tomographic imaging experiments. Although it is already well-known that sinograms posess some amount of redundancy, in this work, we present novel theory suggesting that sinograms will often possess substantial additional redundancies that have not been explicitly exploited by previous methods. Specifically, we derive that sinograms will often satisfy multiple simple data-dependent autoregression relationships. This kind of autoregressive structure enables missing/degraded sinogram samples to be linearly predicted using a simple shift-invariant linear combination of neighboring samples. Our theory also further implies that if sinogram samples are assembled into a structured Hankel/Toeplitz matrix, then the matrix will be expected to have low-rank characteristics. As a result, sinogram restoration problems can be formulated as structured low-rank matrix recovery problems. Illustrations of this approach are provided using several different (real and simulated) X-ray imaging datasets, including comparisons against a state-of-the-art deep learning approach. Results suggest that structured low-rank matrix methods for sinogram recovery can have comparable performance to state-of-the-art approaches. Although our evaluation focuses on competitive comparisons against other approaches, we believe that autoregressive constraints are actually complementary to existing approaches with strong potential synergies.

Poisson-Gaussian Noise Reduction for X-Ray Images Based on Local Linear Minimum Mean Square Error Shrinkage in Nonsubsampled Contourlet Transform Domain

Noise reduction is important for X-ray images because it can reduce radiation exposure to patients. X-ray image noise has a Poisson-Gaussian distribution, and recently, noise analysis and removal in multiscale transformations have been widely implemented. The nonsubsampled contourlet transform (NSCT) is a multiscale transformation suitable for medical images that separates the scale and direction. This study proposes a Poisson-Gaussian noise-removal method using NSCT shrinkage that is based on the characteristics of Poisson-Gaussian noise in NSCT domain. It has the structure of a block-matching 3D filtering algorithm in the form of basic estimation and noise removal process; however, the main processes are modified to consider Poisson-Gaussian noise characteristics. In the basic estimation process, an NSCT shrinkage method that is suitable for Poisson-Gaussian noise characteristics is developed by optimizing the local linear minimum mean square error estimator in the NSCT domain. In the denoising step, the noise term of the Wiener filter is determined using the result of the NSCT shrinkage, and finally, the denoised image is obtained. The proposed method is applied to simulated and real X-ray images and is compared with other state-of-the-art Poisson-Gaussian noise removal methods; it exhibits excellent results in both quantitative and qualitative aspects.

Mammography and breast tomosynthesis simulator for virtual clinical trials

Computer modeling and simulations are increasingly being used to predict the clinical performance of x-ray imaging devices in silico, and to generate synthetic patient images for training and testing of machine learning algorithms. We present a detailed description of the computational models implemented in the open source GPU-accelerated Monte Carlo x-ray imaging simulation code MC-GPU. This code, originally developed to simulate radiography and computed tomography, has been extended to replicate a commercial full-field digital mammography and digital breast tomosynthesis (DBT) device. The code was recently used to image 3000 virtual breast models with the aim of reproducing in silico a clinical trial used in support of the regulatory approval of DBT as a replacement of mammography for breast cancer screening. The updated code implements a more realistic x-ray source model (extended 3D focal spot, tomosynthesis acquisition trajectory, tube motion blurring) and an improved detector model (direct-conversion Selenium detector with depth-of-interaction effects, fluorescence tracking, electronic noise and anti-scatter grid). The software uses a high resolution voxelized geometry model to represent the breast anatomy. To reduce the GPU memory requirements, the code stores the voxels in memory within a binary tree structure. The binary tree is an efficient compression mechanism because many voxels with the same composition are combined in common tree branches while preserving random access to the phantom composition at any location. A delta scattering ray-tracing algorithm which does not require computing ray-voxel interfaces is used to minimize memory access. Multiple software verification and validation steps intended to establish the credibility of the implemented computational models are reported. The software verification was done using a digital quality control phantom and an ideal pinhole camera. The validation was performed reproducing standard bench testing experiments used in clinical practice and comparing with experimental measurements. A sensitivity study intended to assess the robustness of the simulated results to variations in some of the input parameters was performed using an in silico clinical trial pipeline with simulated lesions and mathematical observers. We show that MC-GPU is able to simulate x-ray projections that incorporate many of the sources of variability found in clinical images, and that the simulated results are robust to some uncertainty in the input parameters. Limitations of the implemented computational models are discussed. Program summaryProgram title: MCGPU_VICTRECPC Library link to program files:https://doi.org/10.17632/k5x2bsf27m.1Licensing provisions: CC0 1.0Programming language: C (with NVIDIA CUDA extensions)Nature of problem: The health risks associated with ionizing radiation impose a limit to the amount of clinical testing that can be done with x-ray imaging devices. In addition, radiation dose cannot be directly measured inside the body. For these reasons, a computational replica of an x-ray imaging device that simulates radiographic images of synthetic anatomical phantoms is of great value for device evaluation. The simulated radiographs and dosimetric estimates can be used for system design and optimization, task-based evaluation of image quality, machine learning software training, and in silico imaging trials.Solution method: Computational models of a mammography x-ray source and detector have been implemented. X-ray transport through matter is simulated using Monte Carlo methods customized for parallel execution in multiple Graphics Processing Units. The input patient anatomy is represented by voxels, which are efficiently stored in the video memory using a new binary tree structure compression mechanism.

The effect of artificial X-rays on C-arm positioning performance in a simulated orthopaedic surgical setting.

We designed an Artificial X-ray Imaging System (AXIS) that generates simulated fluoroscopic X-ray images on the fly and assessed its utility in improving C-arm positioning performance by C-arm users with little or no C-arm experience. The AXIS system was comprised of an optical tracking system to monitor C-arm movement, a manikin, a reference CT volume registered to the manikin, and a Digitally Reconstructed Radiograph algorithm to generate live simulated fluoroscopic images. A user study was conducted with 30 participants who had little or no C-arm experience. Each participant carried out four tasks using a real C-arm: an introduction session, an AXIS-guided set of pelvic imaging tasks, a non-AXIS guided set of pelvic imaging tasks, and a questionnaire. For each imaging task, the participant replicated a set of three target X-ray images by taking real radiographs of a manikin with a C-arm. The number of X-rays required, task time, and C-arm positioning accuracy were recorded. We found a significant 53% decrease in the number of X-rays used and a moderate 10-26% improvement in lateral C-arm axis positioning accuracy without requiring more time to complete the tasks when the participants were guided by artificial X-rays. The questionnaires showed that the participants felt significantly more confident in their C-arm positioning ability when they were guided by AXIS. They rated the usefulness of AXIS as very good to excellent, and the realism and accuracy of AXIS as good to very good. Novice users working with a C-arm machine supplemented with the ability to generate simulated X-ray images could successfully accomplish positioning tasks in a simulated surgical setting using markedly fewer X-ray images than when unassisted. In future work, we plan to determine whether such a system can produce similar results in the live operating room without lengthening surgical procedures.

PyNX: high-performance computing toolkit for coherent X-ray imaging based on operators

The open-source PyNX toolkit has been extended to provide tools for coherent X-ray imaging data analysis and simulation. All calculations can be executed on graphical processing units (GPUs) to achieve high-performance computing speeds. The toolkit can be used for coherent diffraction imaging (CDI), ptychography and wavefront propagation, in the far- or near-field regime. Moreover, all imaging operations (propagation, projections, algorithm cycles…) can be implemented in Python as simple mathematical operators, an approach which can be used to easily combine basic algorithms in a tailored chain. Calculations can also be distributed to multiple GPUs, e.g. for large ptychography data sets. Command-line scripts are available for on-line CDI and ptychography analysis, either from raw beamline data sets or using the coherent X-ray imaging data format.

Towards automated in vivo tracheal mucociliary transport measurement: Detecting and tracking particle movement in synchrotron phase-contrast x-ray images

Accurate in vivo quantification of airway mucociliary transport (MCT) in animal models is important for understanding diseases such as cystic fibrosis, as well as for developing therapies. A non-invasive method of measuring MCT behaviour, based on tracking the position of micron sized particles using synchrotron x-ray imaging, has previously been described. In previous studies, the location (and path) of each particle was tracked manually, which is a time consuming and subjective process.Here we describe particle tracking methods that were developed to reduce the need for manual particle tracking. The MCT marker particles were detected in the synchrotron x-ray images using cascade classifiers. The particle trajectories along the airway surface were generated by linking the detected locations between frames using a modified particle linking algorithm. The developed methods were compared with the manual tracking method on simulated x-ray images, as well as on in vivo images of rat airways acquired at the SPring-8 Synchrotron.The results for the simulated and in vivo images showed that the semi-automatic algorithm reduced the time required for particle tracking when compared with the manual tracking method, and was able to detect MCT marker particle locations and measure particle speeds more accurately than the manual tracking method. Future work will examine the modification of methods to improve particle detection and particle linking algorithms to allow for more accurate fully-automatic particle tracking.

Dose image prediction for range and width verifications from carbon ion-induced secondary electron bremsstrahlung x-rays using deep learning workflow.

Imaging of the secondary electron bremsstrahlung (SEB) x rays emitted during particle-ion irradiation is a promising method for beam range estimation. However, the SEB x-ray images are not directly correlated to the dose images. In addition, limited spatial resolution of the x-ray camera and low-count situation may impede correctly estimating the beam range and width in SEB x-ray images. To overcome these limitations of the SEB x-ray images measured by the x-ray camera, a deep learning (DL) approach was proposed in this work to predict the dose images for estimating the range and width of the carbon ion beam on the measured SEB x-ray images. To prepare enough data for the DL training efficiently, 10,000 simulated SEB x-ray and dose image pairs were generated by our in-house developed model function for different carbon ion beam energies and doses. The proposed DL neural network consists of two U-nets for SEB x ray to dose image conversion and super resolution. After the network being trained with these simulated x-ray and dose image pairs, the dose images were predicted from simulated and measured SEB x-ray testing images for performance evaluation. For the 500 simulated testing images, the average mean squared error (MSE) was 2.5×10-5 and average structural similarity index (SSIM) was 0.997 while the error of both beam range and width was within 1mm FWHM. For the three measured SEB x-ray images, the MSE was no worse than 5.5×10-3 and SSIM was no worse than 0.980 while the error of the beam range and width was 2mm and 5mm FWHM, respectively. We have demonstrated the advantages of predicting dose images from not only simulated data but also measured data using our deep learning approach.

Real-space imaging of confined magnetic skyrmion tubes

Magnetic skyrmions are topologically nontrivial particles with a potential application as information elements in future spintronic device architectures. While they are commonly portrayed as two dimensional objects, in reality magnetic skyrmions are thought to exist as elongated, tube-like objects extending through the thickness of the host material. The study of this skyrmion tube state (SkT) is vital for furthering the understanding of skyrmion formation and dynamics for future applications. However, direct experimental imaging of skyrmion tubes has yet to be reported. Here, we demonstrate the real-space observation of skyrmion tubes in a lamella of FeGe using resonant magnetic x-ray imaging and comparative micromagnetic simulations, confirming their extended structure. The formation of these structures at the edge of the sample highlights the importance of confinement and edge effects in the stabilisation of the SkT state, opening the door to further investigation into this unexplored dimension of the skyrmion spin texture.

Personalized cardiovascular intervention simulation system

BackgroundThis study proposes a series of geometry and physics modeling methods for personalized cardiovascular intervention procedures, which can be applied to a virtual endovascular simulator. MethodsBased on personalized clinical computed tomography angiography (CTA) data, mesh models of the cardiovascular system were constructed semi-automatically. By coupling 4D magnetic resonance imaging (MRI) sequences corresponding to a complete cardiac cycle with related physics models, a hybrid kinetic model of the cardiovascular system was built to drive kinematics and dynamics simulation. On that basis, the surgical procedures related to intervention instruments were simulated using specially-designed physics models. These models can be solved in real-time; therefore, the complex interactions between blood vessels and instruments can be well simulated. Additionally, X-ray imaging simulation algorithms and realistic rendering algorithms for virtual intervention scenes are also proposed. In particular, instrument tracking hardware with haptic feedback was developed to serve as the interaction interface of real instruments and the virtual intervention system. Finally, a personalized cardiovascular intervention simulation system was developed by integrating the techniques mentioned above. ResultsThis system supported instant modeling and simulation of personalized clinical data and significantly improved the visual and haptic immersions of vascular intervention simulation. ConclusionsIt can be used in teaching basic cardiology and effectively satisfying the demands of intervention training, personalized intervention planning, and rehearsing.

Study of CT image reconstruction algorithm based on high order total variation

The traditional total variation (TV) minimization algorithm is an image reconstruction algorithm based on compressed sensing, which can accurately reconstruct images from sparse data or highly noisy data and has been widely used in low-dose computed tomography (CT). Sometimes it may lead to staircase effect if the reconstructed image has not obvious piecewise constant feature. Recently, researches in the field of image processing suggested that the high order total variation (HOTV) can effectively suppress staircase effect. Whereas the HOTV reconstruction algorithm has not been carried out deeply and extensively in image reconstruction. Herein, we propose a HOTV reconstruction model and design its adaptive steepest descent-projection onto convex sets (ASD-POCS) solving algorithm, and then characterize its reconstruction performance. We construct the second order TV norm using the second order gradient, and design the data fidelity constrained, second order TV minimization model, and derive the its ASD-POCS algorithm. In order to characterize its reconstruction performance, we use the Shepp-Logan simulation phantom of wavy background, the gray-changing simulation phantom and the chest X-ray image simulation phantom to perform reconstructions. Sparse reconstruction results show that compared to the traditional TV algorithm, the HOTV algorithm can effectively suppress the staircase effect and improve the reconstruction accuracy. Noisy-projections reconstruction results show that the traditional TV algorithm and the HOTV algorithm both have good denoising effect but the HOTV algorithm is slightly better for it can better protect the image edge information. The HOTV algorithm is a better reconstruction algorithm than the TV algorithm in image reconstruction if the piecewise constant feature of the reconstructed image is not so obvious but the grayscale fluctuation feature is prominent. The proposed HOTV algorithm can be extended to various CT configurations and other imaging modalities.

Redundancy gain in visual search of simulated X-ray images.

Cancer diagnosis frequently relies on the interpretation of medical images such as chest X-rays and mammography. This process is error prone; misdiagnoses can reach a rate of 15% or higher. Of particular interest are false negatives-tumors that are present but missed. Previous research has identified several perceptual and attentional problems underlying inaccurate perception of these images. But how might these problems be reduced? The psychological literature has shown that presenting multiple, duplicate images can improve performance. Here we explored whether redundant image presentation can improve target detection in simulated X-ray images, by presenting four identical or similar images concurrently. Displays with redundant images, including duplicates of the same image, showed reduced false-negative rates, compared with displays with a single image. This effect held both when the target's prevalence rate was high and when it was low. Eye tracking showed that fixating on two or more images in the redundant condition speeded target detection and prolonged search, and that the latter effect was the key to reducing false negatives. The redundancy gain may result from both perceptual enhancement and an increase in the search quitting threshold.

A Deep Learning-Based Scatter Correction of Simulated X-ray Images

X-ray scattering significantly limits image quality. Conventional strategies for scatter reduction based on physical equipment or measurements inevitably increase the dose to improve the image quality. In addition, scatter reduction based on a computational algorithm could take a large amount of time. We propose a deep learning-based scatter correction method, which adopts a convolutional neural network (CNN) for restoration of degraded images. Because it is hard to obtain real data from an X-ray imaging system for training the network, Monte Carlo (MC) simulation was performed to generate the training data. For simulating X-ray images of a human chest, a cone beam CT (CBCT) was designed and modeled as an example. Then, pairs of simulated images, which correspond to scattered and scatter-free images, respectively, were obtained from the model with different doses. The scatter components, calculated by taking the differences of the pairs, were used as targets to train the weight parameters of the CNN. Compared with the MC-based iterative method, the proposed one shows better results in projected images, with as much as 58.5% reduction in root-mean-square error (RMSE), and 18.1% and 3.4% increases in peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM), on average, respectively.

Enabling machine learning in X-ray-based procedures via realistic simulation of image formation.

Machine learning-based approaches now outperform competing methods in most disciplines relevant to diagnostic radiology. Image-guided procedures, however, have not yet benefited substantially from the advent of deep learning, in particular because images for procedural guidance are not archived and thus unavailable for learning, and even if they were available, annotations would be a severe challenge due to the vast amounts of data. In silico simulation of X-ray images from 3D CT is an interesting alternative to using true clinical radiographs since labeling is comparably easy and potentially readily available. We extend our framework for fast and realistic simulation of fluoroscopy from high-resolution CT, called DeepDRR, with tool modeling capabilities. The framework is publicly available, open source, and tightly integrated with the software platforms native to deep learning, i.e.,Python, PyTorch, and PyCuda. DeepDRR relies on machine learning for material decomposition and scatter estimation in 3D and 2D, respectively, but uses analytic forward projection and noise injection to ensure acceptable computation times. On two X-ray image analysis tasks, namely (1) anatomical landmark detection and (2) segmentation and localization of robot end-effectors, we demonstrate that convolutional neural networks (ConvNets) trained on DeepDRRs generalize well to real data without re-training or domain adaptation. To this end, we use the exact same training protocol to train ConvNets on naïve and DeepDRRs and compare their performance on data of cadaveric specimens acquired using a clinical C-arm X-ray system. Our findings are consistent across both considered tasks. All ConvNets performed similarly well when evaluated on the respective synthetic testing set. However, when applied to real radiographs of cadaveric anatomy, ConvNets trained on DeepDRRs significantly outperformed ConvNets trained on naïve DRRs ([Formula: see text]). Our findings for both tasks are positive and promising. Combined with complementary approaches, such as image style transfer, the proposed framework for fast and realistic simulation of fluoroscopy from CT contributes to promoting the implementation of machine learning in X-ray-guided procedures. This paradigm shift has the potential to revolutionize intra-operative image analysis to simplify surgical workflows.