Synthetic X-ray Research Articles

Fast X-ray/CT image registration based on perspective projection triangular features

X-ray/CT image registration plays a pivotal role in enhancing surgical navigation success rates. However, challenges stemming from sparse and noisy X-ray image features, coupled with the complexities of multi-parameter optimization, impose limitations on existing methods in terms of registration accuracy and efficiency. In response, this paper presents an innovative approach—a fast X-ray/CT image registration method based on perspective projection triangular features(F-PPTF). By leveraging the conformal nature of perspective projection, the proposed method constructs perspective projection triangular features with rotation, translation, and scale invariance using point feature descriptors. Diverging from multi-parameter iterative optimization techniques, this approach achieves the decoupling of the six transformation parameters. This decoupling simplifies computational intricacies, thereby facilitating swift registration. Experimental evaluations conducted on synthetic and real X-ray images reveal an average rotational absolute error of 0.41°, an average translational absolute error of 1.16 mm, and an average registration time of 14.89 s. In comparison to conventional registration methodologies, the method presented in this paper demonstrates pronounced superiority in terms of both registration accuracy and efficiency, thereby exhibiting heightened potential for broader applicability.

Denoising diffusion probabilistic models for addressing data limitations in chest X-ray classification

Deep learning plays a crucial role in medical imaging analysis, particularly in tasks such as image classification and segmentation. However, learning from medical imaging datasets presents challenges, including scarcity of labeled examples, class imbalances, and inadequate representation of diverse patient populations. To address these challenges, there has been a growing interest in the use of deep generative models to create synthetic training data, with denoising diffusion probabilistic models (DDPMs) recently gaining attention for their ability to produce realistic and high-quality images. This study explores the potential of a DDPM to generate synthetic chest X-rays for multi-label classifier training. The results indicate that the use of a conditional DDPM has the potential to produce a realistic training set of synthetic chest X-rays. In addition, the study analyzes the impact on classification performance of addressing class imbalance. Balancing the synthetic training set increased the overall classification sensitivity from 0.02 to 0.59, but decreased the overall specificity from 0.99 to 0.71. Furthermore, we investigated the potential of unconditional pre-training to learn general representations, followed by conditional fine-tuning of the DDPM. The results indicate that this approach allows the amount of labeled training data to be reduced to 25% of the original set. Finally, we demonstrate that fidelity and classification metrics do not consistently exhibit the same trends. Integrating a DDPM into the classification pipeline underscores the benefits of having optimal control over the data and efficient use of available unlabeled data. Our research provides insights for making informed decisions about integrating generative models into medical image analysis.

Introducing SPINE: A Holistic Approach to Synthetic Pulmonary Imaging Evaluation Through End-to-End Data and Model Management.

In the evolving field of medical imaging and machine learning (ML), this paper introduces a novel framework for evaluating synthetic pulmonary imaging aiming to assess synthetic data quality and applicability. Our study concentrates on synthetic X-ray chest images, crucial for diagnosing respiratory diseases. We employ SPINE (Synthetic Pulmonary Imaging Evaluation) framework, a threefold synthetic images evaluation method including expert domain assessment, statistical data analysis and adversarial evaluation. In order to replicate and validate our methodology, we followed an End-to-End data and model management process which begins with a dataset of Normal and Pneumonia chest X-rays, generating synthetic images using Generative Adversarial Networks (GANs) and training a baseline classifier, essential in the adversarial evaluation axis, testing synthetic images against real data assessing their predictive value. The critical outcome of our approach is the post-market analysis of synthetic images. This innovative method evaluates synthetic images using clinical, statistical, and scientific criteria independently from traditional generation performance metrics. This independent evaluation provides deep insights into the clinical and research effectiveness of the synthetic data. By ensuring these images mirror real data's statistical properties and maintain clinical accuracy, our framework establishes a new standard for the ethical and reliable use of synthetic data in medical imaging and research.

Synthetic 3D Spinal Vertebrae Reconstruction from Biplanar X-rays Utilizing Generative Adversarial Networks.

Computed tomography (CT) offers detailed insights into the internal anatomy of patients, particularly for spinal vertebrae examination. However, CT scans are associated with higher radiation exposure and cost compared to conventional X-ray imaging. In this study, we applied a Generative Adversarial Network (GAN) framework to reconstruct 3D spinal vertebrae structures from synthetic biplanar X-ray images, specifically focusing on anterior and lateral views. The synthetic X-ray images were generated using the DRRGenerator module in 3D Slicer by incorporating segmentations of spinal vertebrae in CT scans for the region of interest. This approach leverages a novel feature fusion technique based on X2CT-GAN to combine information from both views and employs a combination of mean squared error (MSE) loss and adversarial loss to train the generator, resulting in high-quality synthetic 3D spinal vertebrae CTs. A total of n = 440 CT data were processed. We evaluated the performance of our model using multiple metrics, including mean absolute error (MAE) (for each slice of the 3D volume (MAE0) and for the entire 3D volume (MAE)), cosine similarity, peak signal-to-noise ratio (PSNR), 3D peak signal-to-noise ratio (PSNR-3D), and structural similarity index (SSIM). The average PSNR was 28.394 dB, PSNR-3D was 27.432, SSIM was 0.468, cosine similarity was 0.484, MAE0 was 0.034, and MAE was 85.359. The results demonstrated the effectiveness of this approach in reconstructing 3D spinal vertebrae structures from biplanar X-rays, although some limitations in accurately capturing the fine bone structures and maintaining the precise morphology of the vertebrae were present. This technique has the potential to enhance the diagnostic capabilities of low-cost X-ray machines while reducing radiation exposure and cost associated with CT scans, paving the way for future applications in spinal imaging and diagnosis.

Synthetic X-ray emission from white dwarf accreting planetary material

ABSTRACT The emission of hard X-rays associated with white dwarfs (WDs) can be generated by the presence of a stellar companion, either by the companion’s coronal emission or by an accretion disc formed by the material stripped from the companion. Recent studies have suggested that a Jupiter-like planet can also be a donor of the material whose accretion on to the WD can generate hard X-rays. We use the guacho code to reproduce the conditions of this WD–planet scenario. With the example of the hard X-ray WD KPD 0005+5106, we explore different terminal wind velocities and mass-loss rates of a donor planet for a future network of simulations to investigate the luminosity and the spectral and temporal properties of the hard X-ray emission in the WD–planet systems. Our simulations show that the material stripped from the planet forms a disc and accretes on to the WD to reach temperatures high enough to generate hard X-rays as usually seen in X-ray binaries with low-mass companions. For high terminal wind velocities, the planet material does not form a disc, but it rather accretes directly on to the WD surface. The simulations reproduce the X-ray luminosity of another X-ray accreting WD (G 29-38), and only sometimes reaches the hard X-ray luminosity of KPD 0005+5106. The X-ray variability is stochastic and does not reproduce the period of KPD 0005+5106, suggesting that the additional physical processes (e.g. hot spots resulting from magnetic channelling of the accreting material) need to be explored.

Deep learning in computed tomography super resolution using multi-modality data training.

One of the limitations in leveraging the potential of artificial intelligence in X-ray imaging is the limited availability of annotated training data. As X-ray and CT shares similar imaging physics, one could achieve cross-domain data sharing, so to generate labeled synthetic X-ray images from annotated CT volumes as digitally reconstructed radiographs (DRRs). To account for the lower resolution of CT and the CT-generated DRRs as compared to the real X-ray images, we propose the use of super-resolution (SR) techniques to enhance the CT resolution before DRRgeneration. As spatial resolution can be defined by the modulation transfer function kernel in CT physics, we propose to train a SR network using paired low-resolution (LR) and high-resolution (HR) images by varying the kernel's shape and cutoff frequency. This is different to previous deep learning-based SR techniques on RGB and medical images which focused on refining the sampling grid. Instead of generating LR images by bicubic interpolation, we aim to create realistic multi-detector CT (MDCT) like LR images from HR cone-beam CT (CBCT)scans. We propose and evaluate the use of a SR U-Net for the mapping between LR and HR CBCT image slices. We reconstructed paired LR and HR training volumes from the same CT scans with small in-plane sampling grid size of . We used the residual U-Net architecture to train two models. SRUN : trained with kernel-based LR images, and SRUN : trained with bicubic downsampled data as baseline. Both models are trained on one CBCT dataset (n = 13 391). The performance of both models was then evaluated on unseen kernel-based and interpolation-based LR CBCT images (n = 10 950), and also on MDCT images (n = 1392). Five-fold cross validation and ablation study were performed to find the optimal hyperparameters. Both SRUN and SRUN models show significant improvements (p-value 0.05) in mean absolute error (MAE), peak signal-to-noise ratio (PSNR) and structural similarity index measures (SSIMs) on unseen CBCT images. Also, the improvement percentages in MAE, PSNR, and SSIM by SRUN is larger than SRUN . For SRUN , MAE is reduced by 14%, and PSNR and SSIMs increased by 6 and 8%, respectively. To conclude, SRUN outperforms SRUN , which the former generates sharper images when tested with kernel-based LR CBCT images as well as cross-modality LR MDCTdata. Our proposed method showed better performance than the baseline interpolation approach on unseen LR CBCT. We showed that the frequency behavior of the used data is important for learning the SR features. Additionally, we showed cross-modality resolution improvements to LR MDCT images. Our approach is, therefore, a first and essential step in enabling realistic high spatial resolution CT-generated DRRs for deep learningtraining.

Simulated non-thermal emission of the supernova remnant G1.9 + 0.3

ABSTRACT Supernova remnants are the nebular leftover of defunct stellar environments, resulting from the interaction between a supernova blastwave and the circumstellar medium shaped by the progenitor throughout its life. They display a large variety of non-spherical morphologies such as ears that shine non-thermally. We have modelled the structure and the non-thermal emission of the supernova remnant G1.9 + 0.3 through 3D magnetohydrodynamic numerical simulations. We propose that the peculiar ear-shaped morphology of this supernova remnant results from the interaction of its blast wave with a magnetized circumstellar medium, which was previously asymmetrically shaped by the past stellar wind emanating from the progenitor star or its stellar companion. We created synthetic non-thermal radio and X-ray maps from our simulated remnant structure, which are in qualitative agreement with observations, forming ears on the polar directions. Our synthetic map study explains the discrepancies between the measured non-thermal radio and X-ray surface brightness distributions assuming that the inverse Compton process produces the observed X-ray emission.

Deciphering the Morphological Origins of X-shaped Radio Galaxies: Numerical Modeling of Backflow versus Jet Reorientation

X-shaped radio galaxies (XRGs) develop when certain extragalactic jets deviate from their propagation path. An asymmetric ambient medium (backflow model) or complex active galactic nucleus activity (jet-reorientation model) enforcing the jet direction to deviate may cause these structures. In this context, the present investigation focuses on the modeling of XRGs by performing 3D relativistic magnetohydrodynamic simulations. We implement different jet-propagation models applying an initially identical jet-ambient medium configuration to understand distinctive features. This study, the first of its kind, demonstrates that all adopted models produce XRGs with notable properties, thereby challenging the notion of a universal model. Jet reorientation naturally explains several contentious properties of XRGs, including wing alignment along the ambient medium’s primary axis, development of collimated lobes, and the formation of noticeably longer wings than active lobes. These XRGs disrupt the cluster medium by generating isotropic shocks and channeling more energy than in the backflow scenario. Our synthetic thermal X-ray maps of the cluster medium reveal four clear elongated cavities associated with the wing-lobe alignment, regardless of projection effects, but they affect their age estimation. We show that the depth and geometric alignment of the evolved cavities may qualify as promising characteristics of XRGs, which may be used to disentangle different formation scenarios.

Identification and characterization of endo-α-, exo-α-, and exo-β-d-arabinofuranosidases degrading lipoarabinomannan and arabinogalactan of mycobacteria

The cell walls of pathogenic and acidophilic bacteria, such as Mycobacterium tuberculosis and Mycobacterium leprae, contain lipoarabinomannan and arabinogalactan. These components are composed of d-arabinose, the enantiomer of the typical l-arabinose found in plants. The unique glycan structures of mycobacteria contribute to their ability to evade mammalian immune responses. In this study, we identified four enzymes (two GH183 endo-d-arabinanases, GH172 exo-α-d-arabinofuranosidase, and GH116 exo-β-d-arabinofuranosidase) from Microbacterium arabinogalactanolyticum. These enzymes completely degraded the complex d-arabinan core structure of lipoarabinomannan and arabinogalactan in a concerted manner. Furthermore, through biochemical characterization using synthetic substrates and X-ray crystallography, we elucidated the mechanisms of substrate recognition and anomer-retaining hydrolysis for the α- and β-d-arabinofuranosidic bonds in both endo- and exo-mode reactions. The discovery of these d-arabinan-degrading enzymes, along with the understanding of their structural basis for substrate specificity, provides valuable resources for investigating the intricate glycan architecture of mycobacterial cell wall polysaccharides and their contribution to pathogenicity.

Joint Bi-Modal Tomographic Reconstruction from Observations with Missing Data

A method for tomographic reconstruction of bi-modal data sets using cross modal regularization was developed. The method constrains the Simultaneous Algebraic Reconstruction Technique by using clustering-based segmented images and quality controlled fused images, to jointly regularize the reconstruction of structures imaged with different methods. The joint regularized reconstruction method was tested using a synthetic X-Ray and neutron data set with a mix of attenuation driven data starvation. Assuming consistent material boundaries in the different imaging modalities, it was shown that improvements in terms of the Feature Similarity Index Measure are made compared to a traditional SART-TV reconstruction. UK Ministry of Defence ©Crown Owned Copyright 2023/AWE

Data Augmentation of X-ray Images for Automatic Cargo Inspection of Nuclear Items.

As part of establishing a management system to prevent the illegal transfer of nuclear items, automatic nuclear item detection technology is required during customs clearance. However, it is challenging to acquire X-ray images of major nuclear items (e.g., nuclear fuel and gas centrifuges) loaded in cargo with which to train a cargo inspection model. In this work, we propose a new means of data augmentation to alleviate the lack of X-ray training data. The proposed augmentation method generates synthetic X-ray images for the training of semantic segmentation models combining the X-ray images of nuclear items and X-ray cargo background images. To evaluate the effectiveness of the proposed data augmentation technique, we trained representative semantic segmentation models and performed extensive experiments to assess its quantitative and qualitative performance capabilities. Our findings show that multiple item insertions to respond to actual X-ray cargo inspection situations and the resulting occlusion expressions significantly affect the performance of the segmentation models. We believe that this augmentation research will enhance automatic cargo inspections to prevent the illegal transfer of nuclear items at airports and ports.

Automated Diagnosis of COVID-19 Using Synthetic Chest X-Ray Images from Generative Adversarial Networks and Blend of Inception-v3 and Vgg-19 Features

Automated Diagnosis of COVID-19 Using Synthetic Chest X-Ray Images from Generative Adversarial Networks and Blend of Inception-v3 and Vgg-19 Features

Quasi-periodic eruptions from impacts between the secondary and a rigidly precessing accretion disc in an extreme mass-ratio inspiral system

X-ray quasi-periodic eruptions (QPEs) represent a recently discovered example of extreme X-ray variability associated with supermassive black holes. These are high-amplitude bursts recurring every few hours that are detected in the soft X-ray band from the nuclei of nearby galaxies whose optical spectra lack the broad emission lines typically observed in unobscured active galaxies. The physical origin of this new X-ray variability phenomenon is still unknown and several theoretical models have been presented. However, no attempt has been made so far to account for the varying QPE recurrence time and luminosity in individual sources, nor for the diversity of the QPE phenomenology in the different known erupters. We present a semi-analytical model based on an extreme mass-ratio inspiral (EMRI) system where the secondary intersects, along its orbit, a rigidly precessing accretion disc surrounding the primary. We assume that QPEs result from emission from an adiabatically expanding, initially optically thick gas cloud expelled from the disc plane at each impact. We produced synthetic X-ray light curves, which we then compared with X-ray data from four QPE sources: GSN 069, eRO-QPE1, eRO-QPE2, and RX J1301.9+2747. Our model aptly reproduces the diversity of QPE properties between the considered objects and it is also able to naturally account for the varying QPE amplitudes and recurrence times in individual sources. Future implementations will enable us to refine the match with the data and to estimate the system parameters precisely, making additional use of multi-epoch QPE data. We briefly discuss the nature of the secondary object, as well as the possible implications of our findings for the EMRI population at large.

Eruption of a Magnetic Flux Rope in a Comprehensive Radiative Magnetohydrodynamic Simulation of Flare-productive Active Regions

Radiative magnetohydrodynamic simulation includes sufficiently realistic physics to allow for the synthesis of remote sensing observables that can be quantitatively compared with observations. We analyze the largest flare in a simulation of the emergence of large flare-productive active regions described by Chen et al. The flare releases 4.5 × 1031 erg of magnetic energy and is accompanied by a spectacular coronal mass ejection. Synthetic soft X-ray flux of this flare reaches M2 class. The eruption reproduces many key features of observed solar eruptions. A preexisting magnetic flux rope is formed along the highly sheared polarity inversion line between a sunspot pair and is covered by an overlying multipole magnetic field. During the eruption, the progenitor flux rope actively reconnects with the canopy field and evolves to the large-scale multithermal flux rope that is observed in the corona. Meanwhile, the magnetic energy released via reconnection is channeled down to the lower atmosphere and gives rise to bright soft X-ray post-flare loops and flare ribbons that reproduce the morphology and dynamic evolution of observed flares. The model helps to shed light on questions of where and when the a flux rope may form and how the magnetic structures in an eruption are related to observable emission properties.

X-ray radiative transfer in full 3D with SKIRT

Context. Models of active galactic nuclei (AGN) suggest that their circumnuclear media are complex with clumps and filaments, while recent observations hint towards polar extended structures of gas and dust, as opposed to the classical torus paradigm. The X-ray band could form an interesting observational window to study these circumnuclear media in great detail. Aims. We want to extend the radiative transfer code SKIRT with the X-ray processes that govern the broad-band X-ray spectra of obscured AGN, to study the structure of AGN circumnuclear media in full 3D, based on their reflected X-ray emission. Methods. We extended the SKIRT code with Compton scattering on free electrons, photo-absorption and fluorescence by cold atomic gas, scattering on bound electrons, and extinction by dust. This includes a novel treatment of extreme-forward scattering by dust, and a detailed description of anomalous Rayleigh scattering. To verify our X-ray implementation, we performed the first dedicated benchmark of X-ray torus models, comparing five X-ray radiative transfer codes. Results. The resulting radiative transfer code covers the X-ray to millimetre wavelength range self-consistently, has all the features of the established SKIRT framework, is publicly available, and is fully optimised to operate in arbitrary 3D geometries. In the X-ray regime, we find an excellent agreement with the simulation results of the MYTORUS and REFLEX codes, which validates our X-ray implementation. We find some discrepancies with other codes, which illustrates the complexity of X-ray radiative transfer and motivates the need for a robust framework that can handle non-linear 3D radiative transfer effects. We illustrate the 3D nature of the code by producing synthetic X-ray images and spectra of clumpy torus models. Conclusions. SKIRT forms a powerful new tool to model circumnuclear media in full 3D, and make predictions for the X-ray band in addition to the dust-dominated infrared-to-UV wavelength range. The new X-ray functionalities of the SKIRT code allow for uncomplicated access to a broad suite of 3D X-ray models for AGN that can easily be tested and modified. This will be particularly useful with the advent of X-ray microcalorimeter observations in the near future.

Three-dimensional, Time-dependent MHD Simulation of Disk–Magnetosphere–Stellar Wind Interaction in a T Tauri, Protoplanetary System

We present a three-dimensional, time-dependent MHD simulation of the short-term interaction between a protoplanetary disk and the stellar corona in a T Tauri system. The simulation includes the stellar magnetic field, self-consistent coronal heating and stellar wind acceleration, and a disk rotating at sub-Keplerian velocity to induce accretion. We find that, initially, as the system relaxes from the assumed initial conditions, the inner part of the disk winds around and moves inward and close to the star as expected. However, the self-consistent coronal heating and stellar wind acceleration build up the original state after some time, significantly pushing the disk out beyond 10R ⋆. After this initial relaxation period, we do not find clear evidence of a strong, steady accretion flow funneled along coronal field lines, but only weak, sporadic accretion. We produce synthetic coronal X-ray line emission light curves, which show flare-like increases that are not correlated with accretion events nor with heating events. These variations in the line emission flux are the result of compression and expansion due to disk–corona pressure variations. Vertical disk evaporation evolves above and below the disk. However, the disk–stellar wind boundary stays quite stable, and any disk material that reaches the stellar wind region is advected out by the stellar wind.

Wearable synthetic leather-based high-performance X-ray shielding materials enabled by the plant polyphenol- and hierarchical structure-facilitated dispersion

Effective protection against X-ray is the premise of utilizing the X-ray, thus it is critical to develop novel X-ray shielding materials with both low density and high X-ray attenuation efficiency. As the even distribution of high-Z element components is of great significance for increasing the attenuation efficiency of X-ray shielding materials, in this study, the microfiber membrane (MFM), a type of synthetic leather featuring hierarchical structure was chosen to provide large surface area for the dispersion of rare earth (RE) element. Meanwhile, plant polyphenol was utilized to achieve the stable loading and uniform dispersion of the Ce or Er into MFM. Benefiting from the assistance of polyphenol and hierarchical structure of MFM, the even dispersion of RE element was successfully realized. The resultant shielding materials displayed approximately 10% superior X-ray attenuation efficiency compared to that without polyphenol, and an averagely 9% increment in X-ray attenuation efficiency than that without hierarchical structure. Moreover, the obtained composite with a thickness of 2.8 mm displayed superior X-ray shielding performance compared to 0.25 mm lead sheet in 16–83 keV and retained an ultralow density of 1.4 g cm–3. Our research results would shed new light on the manufacture of high-performance X-ray shielding materials with excellent X-ray shielding performance.Graphical

X-ray luminosity function of high-mass X-ray binaries: Studying the signatures of different physical processes using detailed binary evolution calculations

Context. Many physical processes taking place during the evolution of binary stellar systems remain poorly understood. The ever-expanding observational sample of X-ray binaries (XRBs) makes them excellent laboratories for constraining binary evolution theory. Such constraints and useful insights can be obtained by studying the effects of various physical assumptions on synthetic X-ray luminosity functions (XLFs) and comparing them with observed XLFs. Aims. In this work we focus on high-mass X-ray binaries (HMXBs) and study the effects on the XLF of various, poorly constrained assumptions regarding physical processes, such as the common-envelope phase, core collapse, and wind-fed accretion. Methods. We used the new binary population synthesis code POSYDON, which employs extensive precomputed grids of detailed stellar structure and binary evolution models, to simulate the entire evolution of binaries. We generated 96 synthetic XRB populations corresponding to different combinations of model assumptions, including different prescriptions for supernova kicks, supernova remnant masses, common-envelope evolution, circularization at the onset of Roche-lobe overflow, and observable wind-fed accretion. Results. The generated HMXB XLFs are feature-rich, deviating from the commonly assumed single power law. We find a break in our synthetic XLF at luminosity ∼1038 erg s−1, similar to observed XLFs. However, we also find a general overabundance of XRBs (up to a factor of ∼10 for certain model parameter combinations) driven primarily by XRBs with black hole accretors. Assumptions about the transient behavior of Be XRBs, asymmetric supernova kicks, and common-envelope physics can significantly affect the shape and normalization of our synthetic XLFs. We find that less well-studied assumptions regarding the circularization of the orbit at the onset of Roche-lobe overflow and criteria for the formation of an X-ray-emitting accretion disk around wind-accreting black holes can also impact our synthetic XLFs and reduce the discrepancy with observations. Conclusions. Our synthetic XLFs do not always agree well with observations, especially at intermediate X-ray luminosities, which is likely due to uncertainties in the adopted physical assumptions. While some model parameters leave distinct imprints on the shape of the synthetic XLFs and can reduce this deviation, others do not have a significant effect overall. Our study reveals the importance of large-scale parameter studies, highlighting the power of XRBs in constraining binary evolution theory.

A novel iterative algorithm to improve segmentations with deep convolutional neural networks trained with synthetic X-ray computed tomography data (i.S.Sy.Da.T.A)

We propose a novel iterative segmentation algorithm (i.S.Sy.Da.T.A: Iterative Segmentation Synthetic Data Training Algorithm) employing Deep Convolutional Neural Networks and synthetic training data for X-ray tomographic reconstructions of complex microstructures. In our method, we reinforce the synthetic training data with experimental XCT datasets that were automatically segmented in the previous iteration. This strategy produces better segmentations in successive iterations. We test our algorithm with experimental XCT reconstructions of a 6-phase Al-Si Matrix Composite reinforced with ceramic fibers and particles. We perform the analysis in 3D with a special network architecture that demonstrates good generalization with synthetic training data. We show that our iterative algorithm returns better segmentations compared to the standard single training approach. More specifically, phases possessing similar attenuation coefficients can be better segmented: for Al2O3 fibers, SiC particles, and Intermetallics, we see an increase of the Dice score with respect to the classic approach: from 0.49 to 0.54, from 0.66 to 0.72, and from 0.55 to 0.66 respectively. Furthermore, the overall Dice score increases from 0.77 to 0.79. The methods presented in this work are also applicable to other materials and imaging techniques.

Real-time deep neural network-based automatic bowel gas segmentation on X-ray images for particle beam treatment.

Since particle beam distribution is vulnerable to change in bowel gas because of its low density, we developed a deep neural network (DNN) for bowel gas segmentation on X-ray images.We used 6688 image datasets from 209 cases as training data, 736 image datasets from 23 cases as validation data and 102 image datasets from 51 cases as test data (total 283 cases). For the training data, we prepared three types of digitally reconstructed radiographic (DRR) images (all-density, bone and gas) by projecting the treatment planning CT image data. However, the real X-ray images acquired in the treatment room showed low contrast that interfered with manual delineation of bowel gas. Therefore, we used synthetic X-ray images converted from DRR images in addition to real X-ray images.We evaluated DNN segmentation accuracy for the synthetic X-ray images using Intersection over Union, recall, precision, and the Dice coefficient, which measured 0.708 ± 0.208, 0.832 ± 0.170, 0.799 ± 0.191, and 0.807 ± 0.178, respectively. The evaluation metrics for the real X-images were less accurate than those for the synthetic X-ray images (0.408 ± 0237, 0.685 ± 0.326, 0.490 ± 0272, and 0.534 ± 0.271, respectively). Computation time was 29.7 ± 1.3 ms/image. Our DNN appears useful in increasing treatment accuracy in particle beam therapy.