Framework for dual-energy-like chest radiography image synthesis from single-energy computed tomography based on cycle-consistent generative adversarial network.

BACKGROUND: Dual-energy (DE) chest radiography (CXR) enables the selective imaging of two relevant materials, namely, soft tissue and bone structures, to better characterize various chest pathologies (i.e., lung nodule, bony lesions, etc.) and potentially improve CXR-based diagnosis. Recently, deep-learning-based image synthesis techniques have attracted considerable attention as alternatives to existing DE methods (i.e., dual-exposure-based and sandwich-detector-based methods) because software-based bone-only and bone-suppression images in CXR could be useful. PURPOSE: The objective of this study was to develop a new framework for DE-like CXR image synthesis from single-energy computed tomography (CT) based on a cycle-consistent generative adversarial network. METHODS: The core techniques of the proposed framework are divided into three categories: (1) data configuration from the generation of pseudo CXR from single energy CT, (2) learning of the developed network architecture using pseudo CXR and pseudo-DE imaging using a single-energy CT, and (3) inference of the trained network on real single-energy CXR. We performed a visual inspection and comparative evaluation using various metrics and introduced a figure of image quality (FIQ) to consider the effects of our framework on the spatial resolution and noise in terms of a single index through various test cases. RESULTS: Our results indicate that the proposed framework is effective and exhibits potential synthetic imaging ability for two relevant materials: soft tissue and bone structures. Its effectiveness was validated, and its ability to overcome the limitations associated with DE imaging techniques (e.g., increase in exposure dose owing to the requirement of two acquisitions, and emphasis on noise characteristics) via an artificial intelligence technique was presented. CONCLUSIONS: The developed framework addresses X-ray dose issues in the field of radiation imaging and enables pseudo-DE imaging with single exposure.

Feasibility study of deep-learning-based bone suppression incorporated with single-energy material decomposition technique in chest X-rays.

OBJECTIVE: To improve the detection of lung abnormalities in chest X-rays by accurately suppressing overlapping bone structures in the lung area. According to literature on missed lung cancer in chest X-rays, such structures are a significant cause of chest-related diagnostic errors. METHODS: This study presents a deep-learning-based bone suppression method where a residual U-Net model is trained for chest X-rays using data set generated from the single-energy material decomposition (SEMD) technique on CT. Synthetic projection images and soft-tissue selective images were obtained from the CT data set via the SEMD, which were then used as the input and label data of the U-Net network. The trained network was tested on synthetic chest X-rays and two real chest radiographs. RESULTS: Bone-suppressed images of the real chest radiographs obtained by the proposed method were similar to the results from the American Association of Physicists in Medicine lung CT data; pulmonary nodules in the soft-tissue selective images appeared more clearly than in the synthetic projection images. The peak signal-to-noise ratio and structural similarity values measured between the output and the corresponding label images were approximately 17.85 and 0.90, respectively. CONCLUSION: The proposed method effectively yielded bone-suppressed chest X-ray images, indicating its clinical usefulness, and it can improve the detection of lung abnormalities in chest X-rays. ADVANCES IN KNOWLEDGE: The idea of using SEMD to obtain large amounts of paired images for deep-learning-based bone suppression algorithms is novel.

Effectiveness of Non-Local Means Algorithm with an Industrial 3 MeV LINAC High-Energy X-Ray System for Non-Destructive Testing.

Industrial high-energy X-ray imaging systems are widely used for non-destructive testing (NDT) to detect defects in the internal structure of objects. Research on X-ray image noise reduction techniques using image processing has been widely conducted with the aim of improving the detection of defects in objects. In this paper, we propose a non-local means (NLM) denoising algorithm to improve the quality of images obtained using an industrial 3 MeV high-energy X-ray imaging system. We acquired X-ray images using various castings and assessed the performance visually and by obtaining the intensity profile, contrast-to-noise ratio, coefficient of variation, and normalized noise power spectrum. Overall, the quality of images processed by the proposed NLM algorithm is superior to those processed by existing algorithms for the acquired casting images. In conclusion, the NLM denoising algorithm offers an efficient and competitive approach to overcome the noise problem in high-energy X-ray imaging systems, and we expect the accompanying image processing software to facilitate and improve image restoration.

Development of a Skewed Pipe Shear Connector for Precast Concrete Structures.

Joint connection methods, such as shear key and loop bar, improve the structural performance of precast concrete structures; consequently, there is usually decreased workability or constructional efficiency. This paper proposes a high-efficiency skewed pipe shear connector. To resist shear and pull-out forces, the proposed connectors are placed diagonally between precast concrete segments and a cast-in-place concrete joint part on a girder. Design variables (such as the pipe diameter, length, and insertion angle) have been examined to investigate the connection performance of the proposed connector. The results of our testing indicate that the skewed pipe shear connectors have 50% higher ductility and a 15% higher ratio of maximum load to yield strength as compared to the corresponding parameters of the loop bar. Finite element analysis was used for validation. The resulting validation indicates that, compared to the loop bar, the skewed pipe shear connector has a higher ultimate shear and pull-out resistance. These results indicate that the skewed pipe shear connector demonstrates more idealized behavior than the loop bar in precast concrete structures.

Comparison study of reconstruction algorithms for prototype digital breast tomosynthesis using various breast phantoms.

Digital breast tomosynthesis (DBT) is a recently developed system for three-dimensional imaging that offers the potential to reduce the false positives of mammography by preventing tissue overlap. Many qualitative evaluations of digital breast tomosynthesis were previously performed by using a phantom with an unrealistic model and with heterogeneous background and noise, which is not representative of real breasts. The purpose of the present work was to compare reconstruction algorithms for DBT by using various breast phantoms; validation was also performed by using patient images. DBT was performed by using a prototype unit that was optimized for very low exposures and rapid readout. Three algorithms were compared: a back-projection (BP) algorithm, a filtered BP (FBP) algorithm, and an iterative expectation maximization (EM) algorithm. To compare the algorithms, three types of breast phantoms (homogeneous background phantom, heterogeneous background phantom, and anthropomorphic breast phantom) were evaluated, and clinical images were also reconstructed by using the different reconstruction algorithms. The in-plane image quality was evaluated based on the line profile and the contrast-to-noise ratio (CNR), and out-of-plane artifacts were evaluated by means of the artifact spread function (ASF). Parenchymal texture features of contrast and homogeneity were computed based on reconstructed images of an anthropomorphic breast phantom. The clinical images were studied to validate the effect of reconstruction algorithms. The results showed that the CNRs of masses reconstructed by using the EM algorithm were slightly higher than those obtained by using the BP algorithm, whereas the FBP algorithm yielded much lower CNR due to its high fluctuations of background noise. The FBP algorithm provides the best conspicuity for larger calcifications by enhancing their contrast and sharpness more than the other algorithms; however, in the case of small-size and low-contrast microcalcifications, the FBP reduced detectability due to its increased noise. The EM algorithm yielded high conspicuity for both microcalcifications and masses and yielded better ASFs in terms of the full width at half maximum. The higher contrast and lower homogeneity in terms of texture analysis were shown in FBP algorithm than in other algorithms. The patient images using the EM algorithm resulted in high visibility of low-contrast mass with clear border. In this study, we compared three reconstruction algorithms by using various kinds of breast phantoms and patient cases. Future work using these algorithms and considering the type of the breast and the acquisition techniques used (e.g., angular range, dose distribution) should include the use of actual patients or patient-like phantoms to increase the potential for practical applications.

Fully iterative scatter corrected digital breast tomosynthesis using GPU-based fast Monte Carlo simulation and composition ratio update.

PURPOSE: In digital breast tomosynthesis (DBT), scatter correction is highly desirable, as it improves image quality at low doses. Because the DBT detector panel is typically stationary during the source rotation, antiscatter grids are not generally compatible with DBT; thus, a software-based scatter correction is required. This work proposes a fully iterative scatter correction method that uses a novel fast Monte Carlo simulation (MCS) with a tissue-composition ratio estimation technique for DBT imaging. METHODS: To apply MCS to scatter estimation, the material composition in each voxel should be known. To overcome the lack of prior accurate knowledge of tissue composition for DBT, a tissue-composition ratio is estimated based on the observation that the breast tissues are principally composed of adipose and glandular tissues. Using this approximation, the composition ratio can be estimated from the reconstructed attenuation coefficients, and the scatter distribution can then be estimated by MCS using the composition ratio. The scatter estimation and image reconstruction procedures can be performed iteratively until an acceptable accuracy is achieved. For practical use, (i) the authors have implemented a fast MCS using a graphics processing unit (GPU), (ii) the MCS is simplified to transport only x-rays in the energy range of 10-50 keV, modeling Rayleigh and Compton scattering and the photoelectric effect using the tissue-composition ratio of adipose and glandular tissues, and (iii) downsampling is used because the scatter distribution varies rather smoothly. RESULTS: The authors have demonstrated that the proposed method can accurately estimate the scatter distribution, and that the contrast-to-noise ratio of the final reconstructed image is significantly improved. The authors validated the performance of the MCS by changing the tissue thickness, composition ratio, and x-ray energy. The authors confirmed that the tissue-composition ratio estimation was quite accurate under a variety of conditions. Our GPU-based fast MCS implementation took approximately 3 s to generate each angular projection for a 6 cm thick breast, which is believed to make this process acceptable for clinical applications. In addition, the clinical preferences of three radiologists were evaluated; the preference for the proposed method compared to the preference for the convolution-based method was statistically meaningful (p < 0.05, McNemar test). CONCLUSIONS: The proposed fully iterative scatter correction method and the GPU-based fast MCS using tissue-composition ratio estimation successfully improved the image quality within a reasonable computational time, which may potentially increase the clinical utility of DBT.

Optimization of configuration parameters in a newly developed digital breast tomosynthesis system.

The purpose of the present work was to investigate the effects of variable projection-view (PV) and angular dose (AD) distributions on the reconstructed image quality for improving microcalcification detection. The PV densities at central and peripheral sites were varied through the distribution of 21 PVs acquired over ± 25° angular range. To vary the AD distribution, 7 PVs in the central region were targeted with two, four and six times the peripheral dose, and the number of central PVs receiving four times the peripheral dose was increased from 3 to 11. The contrast-to-noise ratio (CNR) for in-focus plane quality and the full width at half maximum (FWHM) of artifact spread function (ASF) for resolution in the z-direction were used. Although the ASF improved with increasing PV densities at two peripheral sites, the CNRs were inferior to those obtained with other subsets. With increasing PV density in the central area, the vertical resolution decreased but the CNR increased. Although increasing the central PV or AD concentrations improved image quality, excessive central densities reduced image quality by increasing noise in peripheral views.

Optimization of the key imaging parameters for detection of microcalcifications in a newly developed digital breast tomosynthesis system.

PURPOSE: The purpose of this study was to investigate the effect of different acquisition parameters and to characterize their relationships in order to improve the detection of microcalcifications using digital breast tomosynthesis (DBT). MATERIALS AND METHODS: DBT imaging parameters were optimized using 32 different acquisition sets with 6 angular ranges (± 5°, ± 10°, ± 13°, ± 17°, ± 21°, and ± 25°) and 8 projection views (PVs) (5, 11, 15, 21, 25, 31, 41, and 51 projections). To investigate the effects of variable angular dose distribution, the acquisition sets were evaluated with delivering more dose toward the central views. RESULTS: Our results show that a wide angular range improved the reconstructed image quality in the z-direction. If a large number of projections are acquired, then electronic noise may dominate the contrast-to-noise ratio (CNR) due to reduced radiation dose per projection. With delivering more dose toward the central views, it was found that the vertical resolution was reduced with increasing dose in the central PVs. On the other hand, the CNR clearly increased with increasing concentration of dose distribution in central views. CONCLUSIONS: Although it was found that increasing angular range improved the vertical resolution, it was also found that the image quality of microcalcifications in the in-focus plane did not improve by increasing the noise due to greater effective breast thickness. Angular dose distributions, with more dose delivered to the central views, generally yielded a higher quality factor than uniform dose distributions.