Deep learning-based extended field of view computed tomography image reconstruction: influence of network design on image estimation outside the scan field of view

The problem of data truncation in Computed Tomography (CT) is caused by the missing data when the patient exceeds the Scan Field of View (SFOV) of a CT scanner. The reconstruction of a truncated scan produces severe truncation artifacts both inside and outside the SFOV. We have employed a deep learning-based approach to extend the field of view and suppress truncation artifacts. Thereby, our aim is to generate a good estimate of the real patient data and not to provide a perfect and diagnostic image even in regions beyond the SFOV of the CT scanner. This estimate could then be used as an input to higher order reconstruction algorithms [1]. To evaluate the influence of the network structure and layout on the results, three convolutional neural networks (CNNs), in particular a general CNN called ConvNet, an autoencoder, and the U-Net architecture have been investigated in this paper. Additionally, the impact of L1, L2, structural dissimilarity and perceptual loss functions on the neural network’s learning have been assessed and evaluated. The evaluation of data set comprising 12 truncated test patients demonstrated that the U-Net in combination with the structural dissimilarity loss showed the best performance in terms of image restoration in regions beyond the SFOV of the CT scanner. Moreover, this network produced the best mean absolute error, L1, L2, and structural dissimilarity evaluation measures on the test set compared to other applied networks. Therefore, it is possible to achieve truncation artifact removal using deep learning techniques.

Phase Diagram and Order Reconstruction Modeling for Nematics in Asymmetric π-Cells

Intense electric fields applied to an asymmetric π-cell containing a nematic liquid crystal subjected to strong mechanical stresses induce distortions that are relaxed through a fast-switching mechanism: the order reconstruction transition. Topologically different nematic textures are connected by such a mechanism that is spatially driven by the intensity of the applied electric fields and by the anchoring angles of the nematic molecules on the confining plates of the cell. Using the finite element method, we implemented the moving mesh partial differential equation numerical technique, and we simulated the nematic evolution inside the cell in the context of the Landau–de Gennes order tensor theory. The order dynamics have been well captured, putting in evidence the possible existence of a metastable biaxial state, and a phase diagram of the nematic texture has been built, therefore confirming the appropriateness of the used technique for the study of this type of problem.

Magnetic Field Reconstruction for a Realistic Multi-Point, Multi-Scale Spacecraft Observatory

Future in situ space plasma investigations will likely involve spatially distributed observatories comprised of multiple spacecraft, beyond the four and five spacecraft configurations currently in operation. Inferring the magnetic field structure across the observatory, and not simply at the observation points, is a necessary step towards characterizing fundamental plasma processes using these unique multi-point, multi-scale data sets. We propose improvements upon the classic first-order reconstruction method, as well as a second-order method, utilizing magnetometer measurements from a realistic nine-spacecraft observatory. The improved first-order method, which averages over select ensembles of four spacecraft, reconstructs the magnetic field associated with simple current sheets and numerical simulations of turbulence accurately over larger volumes compared to second-order methods or first-order methods using a single regular tetrahedron. Using this averaging method on data sets with fewer than nine measurement points, the volume of accurate reconstruction compared to a known magnetic vector field improves approximately linearly with the number of measurement points.

Patterns and Reduced-order Reconstruction of Impinging Wiping Jet Pressure Profile Fluctuation Using Proper Orthogonal Decomposition

Wiping jet impingement pressure is important in controlling the coating mass (thickness) and influencing the smoothness of the thin metallic coating produced in continuous galvanizing lines (CGLs). However, the fluctuation of the impingement pressure profile that directly impacts the coating smoothness has not been adequately understood. To study key features of the impingement pressure fluctuation, the instantaneous impingement pressure profiles obtained from Large Eddy Simulations were analyzed using Proper Orthogonal Decomposition (POD). Dominant fluctuation modes of pressure profiles can be differentiated from the energy contents of the modes corresponding to different jet types namely mixing, non-mixing, and transitional mixing jet. The dominant modes of mixing jets in the wiping region contain comparable strength of all modes (flapping, pulsing, and out-of-phase multi pulsing). Non-mixing jets do not show discernable fluctuation modes and transitional mixing jets show pulsing and flapping modes only. Additionally, instantaneous maximum pressure gradient and their location were determined from the reduced-order reconstruction of the pressure profiles. From the analysis, frequency spectra of the magnitude and location fluctuations of the maximum pressure gradients associated with each of the jet types can be clearly distinguished. This is a knowledge that may be helpful for CGL operators in the operation of wiping jets.

EXPRESS: Grouping Effects in Immediate Reconstruction of Order and the Preconditions for Long-term Learning

One commonly acknowledged role of working memory is to set up conditions for new learning. Yet, it has long been understood that there is not a perfect correspondence between conditions leading to good immediate recall from working memory and conditions leading to good delayed recall from long-term memory. Here, in six experiments, we investigated the relation between grouping effects in immediate and delayed reconstruction of order for word lists. There has been a striking absence of tests of grouping effects in long-term memory. In the first four experiments, items within groups are presented concurrently, which encourages associations between items in a group. Despite that presumably favorable situation for group learning, in Experiments 1 and 2 we found effects of grouping only in immediate order reconstruction and not in delayed reconstruction. When more processing time was allowed (Experiments 3 & 4), grouping effects in both immediate and delayed order reconstruction were obtained. Experiment 5 showed that, with items presented one at a time, but with roughly the same amount of processing time and spatial separation as the previous two experiments, grouping effects were obtained neither in immediate order reconstruction nor in delayed reconstruction. However, in Experiment 6 with a more salient manipulation of grouping, effects of grouping were obtained in immediate order reconstruction, but not in delayed reconstruction. In sum, we demonstrated for the first time that there are mechanisms of temporal grouping that assist working memory but are relatively ineffective for long-term learning, in contrast to more effective, concurrent presentation.

Imaging of small penetrable obstacles based on the topological derivative method

PurposeThe purpose of this paper is to present topological derivatives-based reconstruction algorithms to solve an inverse scattering problem for penetrable obstacles.Design/methodology/approachThe method consists in rewriting the inverse reconstruction problem as a topology optimization problem and then to use the concept of topological derivatives to seek a higher-order asymptotic expansion for the topologically perturbed cost functional. Such expansion is truncated and then minimized with respect to the parameters under consideration, which leads to noniterative second-order reconstruction algorithms.FindingsIn this paper, the authors develop two different classes of noniterative second-order reconstruction algorithms that are able to accurately recover the unknown penetrable obstacles from partial measurements of a field generated by incident waves.Originality/valueThe current paper is a pioneer work in developing a reconstruction method entirely based on topological derivatives for solving an inverse scattering problem with penetrable obstacles. Both algorithms proposed here are able to return the number, location and size of multiple hidden and unknown obstacles in just one step. In summary, the main features of these algorithms lie in the fact that they are noniterative and thus, very robust with respect to noisy data as well as independent of initial guesses.

Application of Proper Orthogonal Decomposition to Study Coherent Flow Structures in a Saccular Aneurysm

Aneurysms are localized expansions of weakened blood vessels that can be debilitating or fatal upon rupture. Previous studies have shown that flow in an aneurysm exhibits complex flow structures that are correlated with its inflow conditions. Therefore, the objective of this study was to demonstrate the application of proper orthogonal decomposition (POD) to study the impact of different inflow conditions on energetic flow structures and their temporal behavior in an aneurysm. To achieve this objective, experiments were performed on an idealized rigid sidewall aneurysm model. A piston pump system was used for precise inflow control, i.e., peak Reynolds number (Rep) and Womersley number (α) were varied from 50 to 270 and 2 to 5, respectively. The velocity flow field measurements at the midplane location of the idealized aneurysm model were performed using particle image velocimetry (PIV). The results demonstrate the efficacy of POD in decomposing complex data, and POD was able to capture the energetic flow structures unique to each studied inflow condition. Furthermore, the time-varying coefficient results highlighted the interplay between the coefficients and their corresponding POD modes, which in turn helped explain how POD modes impact certain flow features. The low-order reconstruction results were able to capture the flow evolution and provide information on complex flow in an aneurysm. The POD and low-order reconstruction results also indicated that vortex formation, evolution, and convection varied with an increase in α, while vortex strength and formation of secondary structures were correlated with an increase in Rep.