RESUMO
Iterative reconstruction algorithms are becoming increasingly important in electron tomography of biological samples. These algorithms, however, impose major computational demands. Parallelization must be employed to maintain acceptable running times. Graphics Processing Units (GPUs) have been demonstrated to be highly cost-effective for carrying out these computations with a high degree of parallelism. In a recent paper by Xu et al. (2010), a GPU implementation strategy was presented that obtains a speedup of an order of magnitude over a previously proposed GPU-based electron tomography implementation. In this technical note, we demonstrate that by making alternative design decisions in the GPU implementation, an additional speedup can be obtained, again of an order of magnitude. By carefully considering memory access locality when dividing the workload among blocks of threads, the GPU's cache is used more efficiently, making more effective use of the available memory bandwidth.
Assuntos
Algoritmos , Tomografia com Microscopia Eletrônica/métodos , Processamento de Imagem Assistida por Computador/métodos , Dispositivos de Armazenamento em Computador , Processamento Eletrônico de Dados , Hemocianinas/ultraestrutura , Processamento de Imagem Assistida por Computador/instrumentaçãoRESUMO
Quantitative electron tomography is proposed to characterize porous materials at a nanoscale. To achieve reliable three-dimensional (3D) quantitative information, the influence of missing wedge artifacts and segmentation methods is investigated. We are presenting the "Discrete Algebraic Reconstruction Algorithm" as the most adequate tomography method to measure porosity at the nanoscale. It provides accurate 3D quantitative information, regardless the presence of a missing wedge. As an example, we applied our approach to nanovoids in La2Zr2O7 thin films.
RESUMO
Accurate segmentation of the jaw (i.e., mandible and maxilla) and the teeth in cone beam computed tomography (CBCT) scans is essential for orthodontic diagnosis and treatment planning. Although various (semi)automated methods have been proposed to segment the jaw or the teeth, there is still a lack of fully automated segmentation methods that can simultaneously segment both anatomic structures in CBCT scans (i.e., multiclass segmentation). In this study, we aimed to train and validate a mixed-scale dense (MS-D) convolutional neural network for multiclass segmentation of the jaw, the teeth, and the background in CBCT scans. Thirty CBCT scans were obtained from patients who had undergone orthodontic treatment. Gold standard segmentation labels were manually created by 4 dentists. As a benchmark, we also evaluated MS-D networks that segmented the jaw or the teeth (i.e., binary segmentation). All segmented CBCT scans were converted to virtual 3-dimensional (3D) models. The segmentation performance of all trained MS-D networks was assessed by the Dice similarity coefficient and surface deviation. The CBCT scans segmented by the MS-D network demonstrated a large overlap with the gold standard segmentations (Dice similarity coefficient: 0.934 ± 0.019, jaw; 0.945 ± 0.021, teeth). The MS-D network-based 3D models of the jaw and the teeth showed minor surface deviations when compared with the corresponding gold standard 3D models (0.390 ± 0.093 mm, jaw; 0.204 ± 0.061 mm, teeth). The MS-D network took approximately 25 s to segment 1 CBCT scan, whereas manual segmentation took about 5 h. This study showed that multiclass segmentation of jaw and teeth was accurate and its performance was comparable to binary segmentation. The MS-D network trained for multiclass segmentation would therefore make patient-specific orthodontic treatment more feasible by strongly reducing the time required to segment multiple anatomic structures in CBCT scans.
Assuntos
Aprendizado Profundo , Ortodontia , Tomografia Computadorizada de Feixe Cônico Espiral , Tomografia Computadorizada de Feixe Cônico , Humanos , Processamento de Imagem Assistida por Computador , Redes Neurais de ComputaçãoRESUMO
The field of discrete tomography focuses on the reconstruction of samples that consist of only a few different materials. Ideally, a three-dimensional (3D) reconstruction of such a sample should contain only one grey level for each of the compositions in the sample. By exploiting this property in the reconstruction algorithm, either the quality of the reconstruction can be improved significantly, or the number of required projection images can be reduced. The discrete reconstruction typically contains fewer artifacts and does not have to be segmented, as it already contains one grey level for each composition. Recently, a new algorithm, called discrete algebraic reconstruction technique (DART), has been proposed that can be used effectively on experimental electron tomography datasets. In this paper, we propose discrete tomography as a general reconstruction method for electron tomography in materials science. We describe the basic principles of DART and show that it can be applied successfully to three different types of samples, consisting of embedded ErSi(2) nanocrystals, a carbon nanotube grown from a catalyst particle and a single gold nanoparticle, respectively.
RESUMO
A novel reconstruction procedure is proposed to achieve atomic resolution in electron tomography. The method exploits the fact that crystals are discrete assemblies of atoms (atomicity). This constraint enables us to obtain a three-dimensional (3-D) reconstruction of test structures from less than 10 projections even in the presence of noise and defects. Phase contrast transmission electron microscopy (TEM) images of a gold nanocrystal were simulated in six different zone axes. The discrete number of atoms in every column is determined by application of the channelling theory to reconstructed electron exit waves. The procedure is experimentally validated by experiments with gold samples. Our results show that discrete tomography recovers the shape of the particle as well as the position of its 309 atoms from only three projections. Experiments on a nanocrystal that contains several missing atoms, both on the surface and in the core of the nanocrystal, while considering a high noise level in each simulated image were performed to prove the stability of the approach to reconstruct defects. The algorithm is well capable of handling structural defects in a highly noisy environment, even if this causes atom count "errors" in the projection data.
RESUMO
Tomographic reconstruction algorithms offer a means by which a tilt-series of transmission images can be combined to yield a three dimensional model of the specimen. Conventional reconstruction algorithms assume that the measured signal is a linear projection of some property, typically the density, of the material. Here we report the use of multislice simulations to investigate the extent to which this assumption is met in HAADF-STEM imaging. The use of simulations allows for a systematic survey of a range of materials and microscope parameters to inform optimal experimental design. Using this approach it is demonstrated that the imaging of amorphous materials is in good agreement with the projection assumption in most cases. Images of crystalline specimens taken along zone-axes are found to be poorly suited for conventional linear reconstruction algorithms due to channelling effects which produce enhanced intensities compared with off-axis images, and poor compliance with the projection requirement. Off-axis images are found to be suitable for reconstruction, though they do not strictly meet the linearity requirement in most cases. It is demonstrated that microscope parameters can be selected to yield improved compliance with the projection requirement.
RESUMO
Electron tomography is currently a versatile tool to investigate the connection between the structure and properties of nanomaterials. However, a quantitative interpretation of electron tomography results is still far from straightforward. Especially accurate quantification of pore-space is hampered by artifacts introduced in all steps of the processing chain, i.e., acquisition, reconstruction, segmentation and quantification. Furthermore, most common approaches require subjective manual user input. In this paper, the PORES algorithm "POre REconstruction and Segmentation" is introduced; it is a tailor-made, integral approach, for the reconstruction, segmentation, and quantification of porous nanomaterials. The PORES processing chain starts by calculating a reconstruction with a nanoporous-specific reconstruction algorithm: the Simultaneous Update of Pore Pixels by iterative REconstruction and Simple Segmentation algorithm (SUPPRESS). It classifies the interior region to the pores during reconstruction, while reconstructing the remaining region by reducing the error with respect to the acquired electron microscopy data. The SUPPRESS reconstruction can be directly plugged into the remaining processing chain of the PORES algorithm, resulting in accurate individual pore quantification and full sample pore statistics. The proposed approach was extensively validated on both simulated and experimental data, indicating its ability to generate accurate statistics of nanoporous materials.
RESUMO
In this work, the simultaneous iterative reconstruction technique (SIRT), the total variation minimization (TVM) reconstruction technique and the discrete algebraic reconstruction technique (DART) for electron tomography are compared and the advantages and disadvantages are discussed. Furthermore, we describe how the result of a three dimensional (3D) reconstruction based on TVM can provide objective information that is needed as the input for a DART reconstruction. This approach results in a tomographic reconstruction of which the segmentation is carried out in an objective manner.
RESUMO
Accurate segmentation of nanoparticles within various matrix materials is a difficult problem in electron tomography. Due to artifacts related to image series acquisition and reconstruction, global thresholding of reconstructions computed by established algorithms, such as weighted backprojection or SIRT, may result in unreliable and subjective segmentations. In this paper, we introduce the Partially Discrete Algebraic Reconstruction Technique (PDART) for computing accurate segmentations of dense nanoparticles of constant composition. The particles are segmented directly by the reconstruction algorithm, while the surrounding regions are reconstructed using continuously varying gray levels. As no properties are assumed for the other compositions of the sample, the technique can be applied to any sample where dense nanoparticles must be segmented, regardless of the surrounding compositions. For both experimental and simulated data, it is shown that PDART yields significantly more accurate segmentations than those obtained by optimal global thresholding of the SIRT reconstruction.
RESUMO
PURPOSE: To reduce beam hardening artifacts in CT in case of an unknown x-ray spectrum and unknown material properties. METHODS: The authors assume that the object can be segmented into a few materials with different attenuation coefficients, and parameterize the spectrum using a small number of energy bins. The corresponding unknown spectrum parameters and material attenuation values are estimated by minimizing the difference between the measured sinogram data and a simulated polychromatic sinogram. Three iterative algorithms are derived from this approach: two reconstruction algorithms IGR and IFR, and one sinogram precorrection method ISP. RESULTS: The methods are applied on real x-ray data of a high and a low-contrast phantom. All three methods successfully reduce the cupping artifacts caused by the beam polychromaticity in such a way that the reconstruction of each homogeneous region is to good accuracy homogeneous, even in case the segmentation of the preliminary reconstruction image is poor. In addition, the results show that the three methods tolerate relatively large variations in uniformity within the segments. CONCLUSIONS: We show that even without prior knowledge about materials or spectrum, effective beam hardening correction can be obtained.
Assuntos
Artefatos , Processamento de Imagem Assistida por Computador/métodos , Tomografia Computadorizada por Raios X/métodos , Imagens de Fantasmas , Polimetil MetacrilatoRESUMO
Grey value thresholding is a segmentation technique commonly applied to tomographic image reconstructions. Many procedures have been proposed to optimally select the grey value thresholds based on the tomogram data only (e.g., using the image histogram). In this paper, a projection distance minimization (PDM) method is presented that uses the tomographic projection data to determine optimal thresholds. These thresholds are computed by minimizing the distance between the forward projection of the segmented image and the measured projection data. An important contribution of the current paper is the efficient implementation of the forward projection method, which makes the use of the original projection data as a segmentation criterion feasible. Simulation experiments applied to various phantom images show that our proposed method obtains superior results compared to established histogram-based methods.